API documentation

One key, one base URL. Chat is OpenAI-compatible; image, video and audio use a submit-then-poll task API. Prices are in credits and come from the same catalog the API serves, so this page is always current.

Getting started

Item	Value
Base URL	`https://api2.glbgpt.com/ai-api/open/v1`
Authentication	`Authorization: Bearer sk-glb-…` — create keys under API Keys. `X-API-Key` is accepted too.
Limits	60 requests/min, 5 concurrent chat streams, 10 media tasks in flight, 5 active keys — per account.

quickstart

from openai import OpenAI

client = OpenAI(
    api_key="sk-glb-...",
    base_url="https://api2.glbgpt.com/ai-api/open/v1",
)

r = client.chat.completions.create(
    model="gpt-6-astra",
    messages=[{"role": "user", "content": "Hello"}],
    stream=True,
)
for chunk in r:
    print(chunk.choices[0].delta.content or "", end="")

Chat completions

POST /v1/chat/completions takes the OpenAI request body and returns the OpenAI response, streamed or not. Use any official OpenAI SDK with base_url and the key changed. The model is one of the ids below.

Field	Type	Description
modelrequired	string	A chat model id from GET /v1/models (type=chat), e.g. gpt-6-astra.
messagesrequired	array	OpenAI message objects: system / user / assistant / tool. Text and image_url content parts are forwarded as-is.
stream	boolean	true = Server-Sent Events chat.completion.chunk frames. The last frame before [DONE] carries usage.
max_tokens	integer	Output cap. Reasoning models spend it on thinking first — a very low value can return an empty message with finish_reason=length.
temperature, top_p, stop, presence_penalty, frequency_penalty	—	Forwarded unchanged to the model when it supports them.
tools, tool_choice	array / string	OpenAI function calling. tool_calls come back in the same shape, streamed or not.
response_format	object	json_object / json_schema where the model supports it.
reasoning_effort	string	none \| low \| medium \| high for reasoning models. If you omit it, we apply the model's known-good default.
n	integer	Only 1 is supported.

Not supported (ignored or rejected): logprobs, logit_bias, seed (chat), user, store, metadata (chat). Streaming always ends with a chunk that carries usage followed by data: [DONE]; the non-stream response carries usage as usual. You are billed on usage.prompt_tokens and usage.completion_tokens at the model's per-1K price. Reasoning tokens count as completion tokens.

Tool calling

{
  "model": "gpt-6-astra",
  "messages": [{"role": "user", "content": "What's the weather in Tokyo?"}],
  "tools": [{
    "type": "function",
    "function": {
      "name": "get_weather",
      "parameters": {"type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"]}
    }
  }],
  "tool_choice": "auto"
}

The reply contains message.tool_calls (or streamed delta.tool_calls); answer with a role: "tool" message carrying tool_call_id, exactly like OpenAI.

Chat models & prices

Model	Provider	Input / 1K tokens	Output / 1K tokens
glm-5.3	Zhipu	5 cr	15 cr
glm-5.3-flash	Zhipu	1 cr	1 cr
glm-5.2	Zhipu	5 cr	15 cr
claude-sonnet-5	Anthropic	7 cr	33 cr
claude-opus-5	Anthropic	17 cr	83 cr
gpt-6-astra	OpenAI	23 cr	117 cr
gpt-5.6-sol	OpenAI	17 cr	100 cr
gpt-5.6-luna	OpenAI	1 cr	4 cr
gpt-5.6-terra	OpenAI	7 cr	40 cr
deepseek-v4.1-flash	DeepSeek	1 cr	4 cr
gpt-5.5	OpenAI	9 cr	78 cr
gpt-4o	OpenAI	10 cr	30 cr
deepseek-v4-pro	DeepSeek	4 cr	13 cr
gpt-5.4-nano	OpenAI	1 cr	4 cr
gpt-5.4-mini	OpenAI	3 cr	15 cr
kimi-k3	Moonshot	10 cr	50 cr
qwen3.8-max	Qwen	7 cr	20 cr
qwen3.8-flash	Qwen	1 cr	2 cr
gemini-3.5-flash-lite	Google	1 cr	8 cr
gemini-3.8-flash	Google	5 cr	25 cr
claude-opus-4.8	Anthropic	17 cr	83 cr
claude-fable-5	Anthropic	33 cr	167 cr
claude-fable-5.1	Anthropic	33 cr	167 cr
claude-opus-4.7	Anthropic	17 cr	83 cr
claude-opus-4.5	Anthropic	17 cr	83 cr
claude-opus-4.6	Anthropic	17 cr	83 cr
claude-haiku-4.5	Anthropic	3 cr	17 cr
gemini-3.1-pro	Google	10 cr	50 cr
gemini-3.6-flash	Google	5 cr	25 cr
gemini-3-pro	Google	13 cr	60 cr
gemini-omni-flash-preview	Google	5 cr	30 cr
gemini-3-flash	Google	2 cr	10 cr
gemini-3.1-flash-lite	Google	1 cr	5 cr
gemini-2.5-flash	Google	1 cr	8 cr
gemini-2.5-pro	Google	8 cr	50 cr
grok-4.3	xAI	4 cr	8 cr
grok-4.6	xAI	7 cr	20 cr
kimi-k2.6	Moonshot	3 cr	13 cr
muse-spark-1.2	Meta	4 cr	14 cr
muse-spark-1.3	Meta	4 cr	14 cr

Media tasks (image, video, audio)

Media generation is asynchronous. Submit a task, poll it until it is succeeded or failed, then download output.url. Credits are reserved when you submit and released if the task fails. Outputs stay available for 30 days.

Method	Path	Purpose
POST /v1/tasks	Submit an image / video / audio task. Credits are reserved now, settled when the task finishes.
GET /v1/tasks/{id}	Task status and, once succeeded, the output URL.
GET /v1/tasks?limit=&after=&status=	Your recent tasks, newest first, cursor paging via after=<id>.
POST /v1/tasks/estimate	Same body as submit; returns the billing tier and credits without creating a task.
GET /v1/models?type=image\|video\|audio	Every media model with price, tiers (video) and accepted parameters.

Request body — common fields

Every model accepts model and prompt; the rest depends on the model — each model's section below lists exactly what it takes, and GET /v1/models/{id} returns the same in parameters. Fields a model doesn't accept are dropped, not rejected, and echoed back in ignored_parameters so you can spot typos.

Field	Type	Description
modelrequired	string	Media model id, e.g. kling-v3, gpt-image-2.5, eleven-v3.
prompt	string	Text prompt. Required for most models; optional for the image tools (remove-bg, face-swap…). Max 20,000 characters, some models cap lower.
images	string[]	Public https URLs or data: URIs. Order matters where two are needed (face-swap: [face, target]; virtual-tryon: [person, garment]). A single image may also be sent as image.
video_url	string	Public https URL of a source / reference video, for models that accept one.
aspect_ratio	string	e.g. 1:1, 16:9, 9:16, 4:3, 3:4, 21:9 — the set a model accepts follows the provider; unsupported values are rejected by the provider.
resolution	string	One of the model's listed values (case-insensitive aliases accepted). Changes the price tier.
duration	integer	Seconds, within the model's range or set. Changes the price tier. Ignored for fixed-length models.
seed	integer	Deterministic sampling where the provider supports it.
negative_prompt	string	What to avoid, for models that support it.
watermark	boolean	Provider watermark on/off where offered.
audio	boolean	Generate audio with the video (kling-v3, kling-v2-6, wan…). Selects the audio price tier.
draft	boolean	flux-3-video: cheaper draft render (hd only).
metadata	object	Model-specific extras (see each model's metadata keys). Unknown keys are dropped and reported in ignored_parameters.
Idempotency-Key (header)	string	Retry the same submit safely: same key + same body returns the original task (Idempotent-Replayed: true); same key + different body → 422.

tasks

# submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-image-2.5", "prompt": "a red bicycle by a white wall", "aspect_ratio": "1:1"}'

# poll until status is "succeeded" — output.url is the image
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Lifecycle & polling

status	Meaning
queued	Accepted, credits reserved, waiting for a worker.
processing	Sent to the provider (or being retried after a transient provider error).
succeeded	output.url is ready. Charged the reserved amount (audio models: settled on real length).
failed	error.code says why. Reserved credits are released.

Poll GET /v1/tasks/{id} every 5–10 seconds. Typical times: images 10–60 s, audio 20–90 s, video 1–5 minutes. A task object looks like this:

{
  "id": "1169054296717658112",
  "object": "task",
  "model": "kling-v3",
  "type": "video",
  "status": "succeeded",
  "created_at": 1789473066,
  "finished_at": 1789473241,
  "credits_reserved": 500,
  "parameters": {"resolution": "720p", "duration": 3},
  "output": {"url": "https://static.futureshareai.com/open/video/…/….mp4", "type": "video"}
}

// failed
{ "id": "…", "status": "failed",
  "error": {"code": "invalid_input", "message": "The provider rejected the input …", "retryable": false} }

error.code	retryable	Meaning
invalid_input	no	Provider rejected the input — unreachable media URL, unsupported combination, or an input without the expected content (e.g. no face detected).
content_policy_violation	no	Blocked by the provider's safety policy or a real-person likeness rule.
rate_limited	yes	Provider busy. Retry in a moment.
provider_unavailable	yes	Provider outage after our retries (4 attempts over ~15 minutes). Resubmit later.
generation_failed	no	Anything else. Not charged.

Estimate before you submit

POST /v1/tasks/estimate with the same body returns the billing tier and credits ({"credits": 500, "billing_model": "kling-v3-720p-3"}) without creating a task or reserving anything.

Image models

Text-to-image models take a prompt and usually an aspect_ratio. Image-editing tools take one or two input images and an optional prompt. Output is a single image URL.

OpenAI · 2

gpt-image-2

OpenAI image generation and editing. Send images for edit / reference; the output is one image.

Inputs: prompt required · images optional (max 4)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
images	string[]	—	Up to 4 input image URL(s) for editing or reference.
metadata.background	string	transparent \| opaque \| auto	Background handling (invalid values are dropped).
metadata.output_format	string	png \| jpeg \| webp	Output file format.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "gpt-image-2",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "1:1",
  "metadata": {
    "background": "auto",
    "output_format": "png"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

gpt-image-2.5

OpenAI image generation and editing. Send images for edit / reference; the output is one image.

Inputs: prompt required · images optional (max 4)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
images	string[]	—	Up to 4 input image URL(s) for editing or reference.

Runs on a different line than gpt-image-2: metadata.background / output_format are ignored here.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "gpt-image-2.5",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "1:1"
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

ByteDance · 3

seedream-5.0-lite

ByteDance Seedream text-to-image / image-to-image.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	With aspect_ratio set the image is rendered at that ratio at the chosen tier's total pixels; without it the provider picks the ratio.
resolution	string	2K \| 4K (anything lower resolves to 2K)	Output resolution tier. Changes the price.
images	string[]	—	Up to 10 reference images.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.
metadata.output_format	string	png \| jpeg	Output file format.
metadata.prompt_optimize_mode	string	standard \| fast	How the provider rewrites the prompt.

seed and negative_prompt are not supported by this model (silently ignored).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedream-5.0-lite",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "1:1",
  "watermark": false,
  "metadata": {
    "output_format": "png",
    "prompt_optimize_mode": "standard"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

seedream-5.0-pro

ByteDance Seedream text-to-image / image-to-image.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	With aspect_ratio set the image is rendered at that ratio at the chosen tier's total pixels; without it the provider picks the ratio.
resolution	string	1K \| 2K (no 4K)	Output resolution tier. Changes the price.
images	string[]	—	Up to 10 reference images.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.
metadata.output_format	string	png \| jpeg	Output file format.
metadata.prompt_optimize_mode	string	standard \| fast	How the provider rewrites the prompt.

seed and negative_prompt are not supported by this model (silently ignored).
Interactive edit: put <bbox> / <point> coordinate tags in the prompt alongside images to target a region.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedream-5.0-pro",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "1:1",
  "watermark": false,
  "metadata": {
    "output_format": "png",
    "prompt_optimize_mode": "standard"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

seedream-4.5

ByteDance Seedream text-to-image / image-to-image.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	With aspect_ratio set the image is rendered at that ratio at the chosen tier's total pixels; without it the provider picks the ratio.
resolution	string	2K \| 4K (anything lower resolves to 2K)	Output resolution tier. Changes the price.
images	string[]	—	Up to 10 reference images.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.
metadata.output_format	string	png \| jpeg	Output file format.
metadata.prompt_optimize_mode	string	standard \| fast	How the provider rewrites the prompt.

seed and negative_prompt are not supported by this model (silently ignored).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedream-4.5",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "1:1",
  "watermark": false,
  "metadata": {
    "output_format": "png",
    "prompt_optimize_mode": "standard"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Google · 3

nano-banana

Google Nano Banana (Gemini image). Text-to-image, or editing / multi-reference with images.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	standard set plus 4:5 and 5:4	Output aspect ratio.
images	string[]	—	Up to 10 reference images (provider cap 14).
seed	integer	—	Random seed for reproducible output.

Flat price; no resolution control (output size follows the model default).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "nano-banana",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "1:1",
  "seed": 42
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

nano-banana-2

Google Nano Banana (Gemini image). Text-to-image, or editing / multi-reference with images.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	standard set plus 4:5 and 5:4	Output aspect ratio.
resolution	string	1K \| 0.5K \| 2K \| 4K (default 1K)	Output resolution tier. Changes the price.
images	string[]	—	Up to 10 reference images (provider cap 14).
seed	integer	—	Random seed for reproducible output.

Priced by resolution tier. 9:21 is clamped to 9:16.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "nano-banana-2",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "1K",
  "aspect_ratio": "1:1",
  "seed": 42
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (4)

tier	credits
resolution=1K	300
resolution=0.5K	200
resolution=2K	400
resolution=4K	600

nano-banana-pro

Google Nano Banana (Gemini image). Text-to-image, or editing / multi-reference with images.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	standard set plus 4:5 and 5:4	Output aspect ratio.
resolution	string	1K \| 2K \| 4K (default 1K)	Output resolution tier. Changes the price.
images	string[]	—	Up to 10 reference images (provider cap 14).
seed	integer	—	Random seed for reproducible output.

Priced by resolution tier (4K is the highest).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "nano-banana-pro",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "1K",
  "aspect_ratio": "1:1",
  "seed": 42
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (2)

tier	credits
resolution=1K	500
resolution=4K	1000

Black Forest Labs · 5

flux-dev

Black Forest Labs FLUX text-to-image.

Inputs: prompt required · images optional (max 8)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
resolution	string	0.5K, 1K, 2K, 4K (video-style 480p/720p/1080p are converted)	Image tier, mapped to pixel width × height. Does not change the price.
seed	integer	—	Random seed for reproducible output.
metadata.prompt_upsampling	boolean	—	Let the provider enrich the prompt.
metadata.output_format	string	jpeg \| png \| webp (default jpeg)	Output file format; invalid values are dropped.
metadata.steps	integer	1–50 (clamped)	Diffusion steps.
metadata.guidance	number	1.5–5 (clamped)	Guidance scale.

Flat price per image regardless of parameters.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "flux-dev",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "aspect_ratio": "1:1",
  "seed": 42,
  "metadata": {
    "prompt_upsampling": false,
    "output_format": "png",
    "steps": 28,
    "guidance": 3
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

flux-pro-1.1

Black Forest Labs FLUX text-to-image.

Inputs: prompt required · images optional (max 8)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
resolution	string	0.5K, 1K, 2K, 4K (video-style 480p/720p/1080p are converted)	Image tier, mapped to pixel width × height. Does not change the price.
seed	integer	—	Random seed for reproducible output.
metadata.prompt_upsampling	boolean	—	Let the provider enrich the prompt.
metadata.output_format	string	jpeg \| png \| webp (default jpeg)	Output file format; invalid values are dropped.

Flat price per image regardless of parameters.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "flux-pro-1.1",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "aspect_ratio": "1:1",
  "seed": 42,
  "metadata": {
    "prompt_upsampling": false,
    "output_format": "png"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

flux-pro-1.1-ultra

Black Forest Labs FLUX 1.1 Ultra — highest quality, aspect passed straight through, no resolution tier.

Inputs: prompt required · images optional (max 8)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
seed	integer	—	Random seed for reproducible output.
metadata.prompt_upsampling	boolean	—	Let the provider enrich the prompt.
metadata.output_format	string	jpeg \| png \| webp (default jpeg)	Output file format; invalid values are dropped.
metadata.raw	boolean	—	Less-processed, more photographic look (Ultra's signature option).

Flat price per image regardless of parameters.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "flux-pro-1.1-ultra",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "aspect_ratio": "1:1",
  "seed": 42,
  "metadata": {
    "prompt_upsampling": false,
    "output_format": "png",
    "raw": false
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

flux-2-pro

Black Forest Labs FLUX text-to-image.

Inputs: prompt required · images optional (max 8)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
resolution	string	0.5K, 1K, 2K, 4K (video-style 480p/720p/1080p are converted)	Image tier, mapped to pixel width × height. Does not change the price.
images	string[]	—	Up to 8 reference images (multi-reference).
seed	integer	—	Random seed for reproducible output.
metadata.prompt_upsampling	boolean	—	Let the provider enrich the prompt.
metadata.output_format	string	jpeg \| png \| webp (default jpeg)	Output file format; invalid values are dropped.

Flat price per image regardless of parameters.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "flux-2-pro",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "1:1",
  "seed": 42,
  "metadata": {
    "prompt_upsampling": false,
    "output_format": "png"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

flux-kontext-pro

Black Forest Labs FLUX Kontext — image editing driven by a prompt.

Inputs: prompt required · 1 image required (max 4)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
images	string[]	—	1 to 4 input images (required; multi-reference is experimental upstream).
seed	integer	—	Random seed for reproducible output.
metadata.prompt_upsampling	boolean	—	Let the provider enrich the prompt.
metadata.output_format	string	jpeg \| png \| webp (default png)	Output file format; invalid values are dropped.

Flat price per image regardless of parameters.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "flux-kontext-pro",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "aspect_ratio": "1:1",
  "seed": 42,
  "metadata": {
    "prompt_upsampling": false,
    "output_format": "png"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Ideogram · 3

ideogram-v4

Ideogram 4.0. Text-to-image, styled text-to-image (metadata style keys), or remix with exactly one input image.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
aspect_ratio	string	Ideogram's 15 ratios (see Ideogram 3.0)	Output aspect ratio.
resolution	string	1K \| 2K (default 2K); an exact WxH from Ideogram's official list is also accepted	Output tier. Does not change the price.
images	string[]	—	Exactly one image (≤ 10 MB) switches to remix / image-to-image.
metadata.aesthetics	string	—	Style description — free text. Any style key switches the request to Ideogram's structured prompt and disables magic prompt.
metadata.art_style	string	—	Style description — free text.
metadata.lighting	string	—	Style description — free text.
metadata.medium	string	—	Style description — free text.
metadata.photo	string	—	Style description — free text.
metadata.color_palette	string[]	["#RRGGBB", ...]	Palette as hex strings (malformed entries dropped).
metadata.image_weight	integer	1–100 (clamped)	Remix only: how strongly the input image steers the result.

seed and negative_prompt are ignored (no upstream fields on 4.0).
Style keys are ignored on remix.
Flat price per image.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "ideogram-v4",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "aspect_ratio": "1:1",
  "metadata": {
    "aesthetics": "clean, minimal",
    "art_style": "editorial illustration",
    "lighting": "soft morning light",
    "medium": "digital painting",
    "photo": "35mm film look",
    "image_weight": 50
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

ideogram-v4-turbo

Ideogram 4.0. Text-to-image, styled text-to-image (metadata style keys), or remix with exactly one input image.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
aspect_ratio	string	Ideogram's 15 ratios (see Ideogram 3.0)	Output aspect ratio.
resolution	string	1K \| 2K (default 2K); an exact WxH from Ideogram's official list is also accepted	Output tier. Does not change the price.
images	string[]	—	Exactly one image (≤ 10 MB) switches to remix / image-to-image.
metadata.aesthetics	string	—	Style description — free text. Any style key switches the request to Ideogram's structured prompt and disables magic prompt.
metadata.art_style	string	—	Style description — free text.
metadata.lighting	string	—	Style description — free text.
metadata.medium	string	—	Style description — free text.
metadata.photo	string	—	Style description — free text.
metadata.color_palette	string[]	["#RRGGBB", ...]	Palette as hex strings (malformed entries dropped).
metadata.image_weight	integer	1–100 (clamped)	Remix only: how strongly the input image steers the result.

seed and negative_prompt are ignored (no upstream fields on 4.0).
Style keys are ignored on remix.
Flat price per image.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "ideogram-v4-turbo",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "aspect_ratio": "1:1",
  "metadata": {
    "aesthetics": "clean, minimal",
    "art_style": "editorial illustration",
    "lighting": "soft morning light",
    "medium": "digital painting",
    "photo": "35mm film look",
    "image_weight": 50
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

ideogram-v4-quality

Ideogram 4.0. Text-to-image, styled text-to-image (metadata style keys), or remix with exactly one input image.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
aspect_ratio	string	Ideogram's 15 ratios (see Ideogram 3.0)	Output aspect ratio.
resolution	string	1K \| 2K (default 2K); an exact WxH from Ideogram's official list is also accepted	Output tier. Does not change the price.
images	string[]	—	Exactly one image (≤ 10 MB) switches to remix / image-to-image.
metadata.aesthetics	string	—	Style description — free text. Any style key switches the request to Ideogram's structured prompt and disables magic prompt.
metadata.art_style	string	—	Style description — free text.
metadata.lighting	string	—	Style description — free text.
metadata.medium	string	—	Style description — free text.
metadata.photo	string	—	Style description — free text.
metadata.color_palette	string[]	["#RRGGBB", ...]	Palette as hex strings (malformed entries dropped).
metadata.image_weight	integer	1–100 (clamped)	Remix only: how strongly the input image steers the result.

seed and negative_prompt are ignored (no upstream fields on 4.0).
Style keys are ignored on remix.
Flat price per image.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "ideogram-v4-quality",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "aspect_ratio": "1:1",
  "metadata": {
    "aesthetics": "clean, minimal",
    "art_style": "editorial illustration",
    "lighting": "soft morning light",
    "medium": "digital painting",
    "photo": "35mm film look",
    "image_weight": 50
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Agnes · 1

agnes-image-2.1-flash

Agnes Image 2.1 Flash — text-to-image or image-to-image.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
resolution	string	0.5K, 1K, 2K, 4K (video-style 480p/720p/1080p are converted)	Size is derived from aspect_ratio + resolution (16-aligned); 0.5K is treated as 1K.
images	string[]	—	Reference images for image-to-image.

The provider renders synchronously — the task can take up to ~6 minutes to leave queued.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "agnes-image-2.1-flash",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "1:1"
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Image tools · 8

remove-bg

Remove the background from images[0].

Inputs: prompt optional · 1 image required (max 1)

Field	Type	Values	Description
images	string[]	—	Exactly 1 image (required).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "remove-bg",
  "images": [
    "https://example.com/input-1.jpg"
  ]
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

remove-watermark

Remove a watermark from images[0].

Inputs: prompt optional · 1 image required (max 1)

Field	Type	Values	Description
images	string[]	—	Exactly 1 image (required).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "remove-watermark",
  "images": [
    "https://example.com/input-1.jpg"
  ]
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

ai-expand

Outpaint images[0] beyond its borders; the prompt guides the new areas.

Inputs: prompt optional · 1 image required (max 1)

Field	Type	Values	Description
images	string[]	—	Exactly 1 image (required).
metadata.aspect_ratio	string	—	Output crop / canvas ratio (this tool reads it from metadata, not the top level).
seed	integer	—	Random seed for reproducible output.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "ai-expand",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "seed": 42,
  "metadata": {
    "aspect_ratio": "16:9"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

id-photo

Turn images[0] into an ID / passport photo.

Inputs: prompt optional · 1 image required (max 1)

Field	Type	Values	Description
images	string[]	—	Exactly 1 image (required).
metadata.gender	string	—	Subject gender hint.
metadata.aspect_ratio	string	—	Output crop / canvas ratio (this tool reads it from metadata, not the top level).
metadata.background	string	—	Background colour / style.
metadata.output_format	string	jpg \| png (webp becomes png)	Output file format.
seed	integer	—	Random seed for reproducible output.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "id-photo",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "seed": 42,
  "metadata": {
    "gender": "female",
    "aspect_ratio": "16:9",
    "background": "auto",
    "output_format": "png"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

face-swap

Swap the face from images[0] onto images[1].

Inputs: prompt optional · 2 images required (max 2)

Field	Type	Values	Description
images	string[]	—	Exactly 2 images: [face source, target photo]. Both must contain a detectable face.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "face-swap",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ]
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

change-haircut

Change the haircut / hair colour in images[0].

Inputs: prompt required · 1 image required (max 1)

Field	Type	Values	Description
images	string[]	—	Exactly 1 image (required).
metadata.haircut	string	—	Target haircut name.
metadata.hair_color	string	—	Target hair colour.
metadata.gender	string	—	Subject gender hint.
metadata.aspect_ratio	string	—	Output crop / canvas ratio (this tool reads it from metadata, not the top level).
metadata.output_format	string	jpg \| png (webp becomes png)	Output file format.
seed	integer	—	Random seed for reproducible output.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "change-haircut",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "seed": 42,
  "metadata": {
    "haircut": "bob",
    "hair_color": "auburn",
    "gender": "female",
    "aspect_ratio": "16:9",
    "output_format": "png"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

virtual-tryon

Put the garment from images[1] on the person in images[0].

Inputs: prompt optional · 2 images required (max 2)

Field	Type	Values	Description
images	string[]	—	Exactly 2 images: [person, garment].
metadata.category	string	upper_body \| lower_body \| dresses (default upper_body)	Garment category.
metadata.garment_des	string	defaults to the prompt	Garment description.
seed	integer	—	Random seed for reproducible output.

prompt doubles as the garment description.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "virtual-tryon",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "seed": 42,
  "metadata": {
    "category": "upper_body",
    "garment_des": "red cotton t-shirt"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

change-background

Replace the background of images[0] from a prompt or a reference image.

Inputs: prompt required · 1 image required (max 1)

Field	Type	Values	Description
images	string[]	—	Exactly 1 image (required).
metadata.ref_image_url	string	—	Reference background image (instead of a prompt).
negative_prompt	string	—	What to avoid.
seed	integer	—	Random seed for reproducible output.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "change-background",
  "prompt": "A red bicycle leaning on a white wall, photo",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "negative_prompt": "blurry, low quality, text",
  "seed": 42,
  "metadata": {
    "ref_image_url": "https://example.com/background.jpg"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Topaz · 1

hd-upscale

Upscale images[0] (Topaz Labs, 4× by default).

Inputs: prompt optional · 1 image required (max 1)

Field	Type	Values	Description
images	string[]	—	Exactly 1 image (required).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "hd-upscale",
  "images": [
    "https://example.com/input-1.jpg"
  ]
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Video models

The price is a tier of resolution × duration (plus audio, reference-video or draft variants where offered). Send an image in images for image-to-video; models that accept video_url or reference_videos do video-to-video. Expand a model's price tiers to see every combination.

Kling · 2

kling-v3

Kling V3 text-to-video / image-to-video (1 image = start frame, 2 = start + end).

Inputs: prompt required · images optional (max 2)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
resolution	string	720p \| 1080p (default 720p)	Output resolution tier. Changes the price.
duration	integer	3–15 s (default 5)	Length in seconds. Changes the price.
images	string[]	—	Up to 2 images: start frame, or start + end frame.
negative_prompt	string	—	What to avoid.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.
audio	boolean	—	Generate an audio track. Selects the audio price tier.

Price = resolution tier × duration, plus the audio variant.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kling-v3",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "720p",
  "duration": 5,
  "aspect_ratio": "16:9",
  "negative_prompt": "blurry, low quality, text",
  "watermark": false,
  "audio": true
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (52)

tier	credits
resolution=720p · duration=3	500
resolution=720p · duration=3 · audio	700
resolution=720p · duration=4	600
resolution=720p · duration=4 · audio	900
resolution=720p · duration=5	700
resolution=720p · duration=5 · audio	1100
resolution=720p · duration=6	900
resolution=720p · duration=6 · audio	1300
resolution=720p · duration=7	1000
resolution=720p · duration=7 · audio	1500
resolution=720p · duration=8	1200
resolution=720p · duration=8 · audio	1700
resolution=720p · duration=9	1300
resolution=720p · duration=9 · audio	1900
resolution=720p · duration=10	1400
resolution=720p · duration=10 · audio	2100
resolution=720p · duration=11	1600
resolution=720p · duration=11 · audio	2400
resolution=720p · duration=12	1700
resolution=720p · duration=12 · audio	2600
resolution=720p · duration=13	1900
resolution=720p · duration=13 · audio	2800
resolution=720p · duration=14	2000
resolution=720p · duration=14 · audio	3000
resolution=720p · duration=15	2100
resolution=720p · duration=15 · audio	3200
resolution=1080p · duration=3	600
resolution=1080p · duration=3 · audio	900
resolution=1080p · duration=4	800
resolution=1080p · duration=4 · audio	1200
resolution=1080p · duration=5	1000
resolution=1080p · duration=5 · audio	1400
resolution=1080p · duration=6	1200
resolution=1080p · duration=6 · audio	1700
resolution=1080p · duration=7	1400
resolution=1080p · duration=7 · audio	2000
resolution=1080p · duration=8	1500
resolution=1080p · duration=8 · audio	2300
resolution=1080p · duration=9	1700
resolution=1080p · duration=9 · audio	2600
resolution=1080p · duration=10	1900
resolution=1080p · duration=10 · audio	2800
resolution=1080p · duration=11	2100
resolution=1080p · duration=11 · audio	3100
resolution=1080p · duration=12	2300
resolution=1080p · duration=12 · audio	3400
resolution=1080p · duration=13	2500
resolution=1080p · duration=13 · audio	3700
resolution=1080p · duration=14	2700
resolution=1080p · duration=14 · audio	4000
resolution=1080p · duration=15	2800
resolution=1080p · duration=15 · audio	4200

kling-v2-6

Kling V2.6 text-to-video / image-to-video.

Inputs: prompt required · images optional (max 2)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
duration	integer	5 \| 10 s (default 5)	Length in seconds. Changes the price.
images	string[]	—	Up to 2 images: start frame, or start + end frame.
negative_prompt	string	—	What to avoid.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.
audio	boolean	—	Generate an audio track. Selects the audio price tier.

Price = resolution tier × duration, plus the audio variant.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kling-v2-6",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "duration": 5,
  "aspect_ratio": "16:9",
  "negative_prompt": "blurry, low quality, text",
  "watermark": false,
  "audio": true
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (8)

tier	credits
duration=5	600
duration=5 · audio	1200
duration=5 · image_to_video	600
duration=5 · audio · image_to_video	1200
duration=10	1200
duration=10 · audio	2400
duration=10 · image_to_video	1200
duration=10 · audio · image_to_video	2400

Alibaba · 1

wan3.0-video

Alibaba Wan 3.0 — one model for text-to-video, first/last-frame, and reference generation (from images, videos, audio, a document or a web page). The mode follows the media you send.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
resolution	string	480P \| 720P \| 1080P (default 720P) — (1080P costs 4× 480P)	Output resolution tier. Changes the price.
duration	integer	2–30 s (default 5)	Length in seconds. Changes the price.
aspect_ratio	string	16:9, 4:3, 1:1, 3:4, 9:16, or adaptive	Output aspect ratio.
images	string[]	—	Frames: [0] first frame, [1] last frame. Cannot be mixed with metadata.reference_*.
metadata.reference_images	string[]	≤ 10	Reference images, addressed in the prompt as 图1, 图2…
metadata.reference_videos	string[]	≤ 5, ≤ 15 s total	Reference videos (视频1…).
metadata.reference_audios	string[]	≤ 5, ≤ 15 s total	Reference audio (音频1…).
metadata.reference_files	string[]	exactly 1; exclusive with reference_links	Generate from a document.
metadata.reference_links	string[]	exactly 1; exclusive with reference_files	Generate from a web page.
metadata.generate_audio	boolean	—	Audio track on/off (same price either way).
metadata.prompt_extend	boolean	—	Provider prompt rewrite.
seed	integer	—	Random seed for reproducible output.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.

resolution has no default — omitting it is a 400.
Only output seconds are billed.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "wan3.0-video",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "720P",
  "duration": 5,
  "aspect_ratio": "16:9",
  "seed": 42,
  "watermark": false,
  "metadata": {
    "reference_images": [
      "https://example.com/reference-1.jpg",
      "https://example.com/reference-2.jpg"
    ],
    "reference_videos": [
      "https://example.com/reference.mp4"
    ],
    "reference_audios": [
      "https://example.com/reference.mp3"
    ],
    "reference_files": [
      "https://example.com/brief.pdf"
    ],
    "reference_links": [
      "https://example.com/article"
    ],
    "generate_audio": true,
    "prompt_extend": true
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (87)

tier	credits
resolution=480P · duration=2	200
resolution=480P · duration=3	300
resolution=480P · duration=4	400
resolution=480P · duration=5	500
resolution=480P · duration=6	600
resolution=480P · duration=7	700
resolution=480P · duration=8	800
resolution=480P · duration=9	900
resolution=480P · duration=10	1000
resolution=480P · duration=11	1100
resolution=480P · duration=12	1200
resolution=480P · duration=13	1300
resolution=480P · duration=14	1400
resolution=480P · duration=15	1500
resolution=480P · duration=16	1600
resolution=480P · duration=17	1700
resolution=480P · duration=18	1800
resolution=480P · duration=19	1900
resolution=480P · duration=20	2000
resolution=480P · duration=21	2100
resolution=480P · duration=22	2200
resolution=480P · duration=23	2300
resolution=480P · duration=24	2400
resolution=480P · duration=25	2500
resolution=480P · duration=26	2600
resolution=480P · duration=27	2700
resolution=480P · duration=28	2800
resolution=480P · duration=29	2900
resolution=480P · duration=30	3000
resolution=720P · duration=2	400
resolution=720P · duration=3	600
resolution=720P · duration=4	800
resolution=720P · duration=5	1000
resolution=720P · duration=6	1200
resolution=720P · duration=7	1400
resolution=720P · duration=8	1600
resolution=720P · duration=9	1800
resolution=720P · duration=10	2000
resolution=720P · duration=11	2200
resolution=720P · duration=12	2400
resolution=720P · duration=13	2600
resolution=720P · duration=14	2800
resolution=720P · duration=15	3000
resolution=720P · duration=16	3200
resolution=720P · duration=17	3400
resolution=720P · duration=18	3600
resolution=720P · duration=19	3800
resolution=720P · duration=20	4000
resolution=720P · duration=21	4200
resolution=720P · duration=22	4400
resolution=720P · duration=23	4600
resolution=720P · duration=24	4800
resolution=720P · duration=25	5000
resolution=720P · duration=26	5200
resolution=720P · duration=27	5400
resolution=720P · duration=28	5600
resolution=720P · duration=29	5800
resolution=720P · duration=30	6000
resolution=1080P · duration=2	800
resolution=1080P · duration=3	1200
resolution=1080P · duration=4	1600
resolution=1080P · duration=5	2000
resolution=1080P · duration=6	2400
resolution=1080P · duration=7	2800
resolution=1080P · duration=8	3200
resolution=1080P · duration=9	3600
resolution=1080P · duration=10	4000
resolution=1080P · duration=11	4400
resolution=1080P · duration=12	4800
resolution=1080P · duration=13	5200
resolution=1080P · duration=14	5600
resolution=1080P · duration=15	6000
resolution=1080P · duration=16	6400
resolution=1080P · duration=17	6800
resolution=1080P · duration=18	7200
resolution=1080P · duration=19	7600
resolution=1080P · duration=20	8000
resolution=1080P · duration=21	8400
resolution=1080P · duration=22	8800
resolution=1080P · duration=23	9200
resolution=1080P · duration=24	9600
resolution=1080P · duration=25	10000
resolution=1080P · duration=26	10400
resolution=1080P · duration=27	10800
resolution=1080P · duration=28	11200
resolution=1080P · duration=29	11600
resolution=1080P · duration=30	12000

HappyHorse · 4

happyhorse-1.0-t2v

HappyHorse 1.0 text-to-video.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
aspect_ratio	string	standard set plus 4:5 and 5:4 (default 16:9; 3:2 / 2:3 snap to 4:3 / 3:4)	Output aspect ratio.
resolution	string	720p \| 1080p (default 720p)	Output resolution tier. Changes the price.
duration	integer	3–15 s (default 5)	Length in seconds. Changes the price.
seed	integer	—	Random seed for reproducible output.
watermark	boolean	default false	Provider watermark (always sent; default false).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "happyhorse-1.0-t2v",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "resolution": "720p",
  "duration": 5,
  "aspect_ratio": "16:9",
  "seed": 42,
  "watermark": false
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (26)

tier	credits
resolution=720p · duration=3	1300
resolution=720p · duration=4	1700
resolution=720p · duration=5	2100
resolution=720p · duration=6	2500
resolution=720p · duration=7	2900
resolution=720p · duration=8	3300
resolution=720p · duration=9	3700
resolution=720p · duration=10	4100
resolution=720p · duration=11	4500
resolution=720p · duration=12	4900
resolution=720p · duration=13	5300
resolution=720p · duration=14	5700
resolution=720p · duration=15	6100
resolution=1080p · duration=3	2000
resolution=1080p · duration=4	2600
resolution=1080p · duration=5	3200
resolution=1080p · duration=6	3900
resolution=1080p · duration=7	4500
resolution=1080p · duration=8	5200
resolution=1080p · duration=9	5800
resolution=1080p · duration=10	6400
resolution=1080p · duration=11	7100
resolution=1080p · duration=12	7700
resolution=1080p · duration=13	8400
resolution=1080p · duration=14	9000
resolution=1080p · duration=15	9600

happyhorse-1.0-i2v

HappyHorse 1.0 image-to-video (aspect follows the input image).

Inputs: prompt required · 1 image required (max 1)

Field	Type	Values	Description
resolution	string	720p \| 1080p (default 720p)	Output resolution tier. Changes the price.
duration	integer	3–15 s (default 5)	Length in seconds. Changes the price.
images	string[]	—	Required: 1 first-frame image.
seed	integer	—	Random seed for reproducible output.
watermark	boolean	default false	Provider watermark (always sent; default false).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "happyhorse-1.0-i2v",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "resolution": "720p",
  "duration": 5,
  "seed": 42,
  "watermark": false
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (26)

tier	credits
resolution=720p · duration=3	1300
resolution=720p · duration=4	1700
resolution=720p · duration=5	2100
resolution=720p · duration=6	2500
resolution=720p · duration=7	2900
resolution=720p · duration=8	3300
resolution=720p · duration=9	3700
resolution=720p · duration=10	4100
resolution=720p · duration=11	4500
resolution=720p · duration=12	4900
resolution=720p · duration=13	5300
resolution=720p · duration=14	5700
resolution=720p · duration=15	6100
resolution=1080p · duration=3	2000
resolution=1080p · duration=4	2600
resolution=1080p · duration=5	3200
resolution=1080p · duration=6	3900
resolution=1080p · duration=7	4500
resolution=1080p · duration=8	5200
resolution=1080p · duration=9	5800
resolution=1080p · duration=10	6400
resolution=1080p · duration=11	7100
resolution=1080p · duration=12	7700
resolution=1080p · duration=13	8400
resolution=1080p · duration=14	9000
resolution=1080p · duration=15	9600

happyhorse-1.0-r2v

HappyHorse 1.0 reference-to-video (1–9 reference images).

Inputs: prompt required · 1 image required (max 9)

Field	Type	Values	Description
aspect_ratio	string	standard set plus 4:5 and 5:4 (default 16:9; 3:2 / 2:3 snap to 4:3 / 3:4)	Output aspect ratio.
resolution	string	720p \| 1080p (default 720p)	Output resolution tier. Changes the price.
duration	integer	3–15 s (default 5)	Length in seconds. Changes the price.
images	string[]	—	Required: 1–9 reference images.
seed	integer	—	Random seed for reproducible output.
watermark	boolean	default false	Provider watermark (always sent; default false).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "happyhorse-1.0-r2v",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "resolution": "720p",
  "duration": 5,
  "aspect_ratio": "16:9",
  "seed": 42,
  "watermark": false
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (26)

tier	credits
resolution=720p · duration=3	1300
resolution=720p · duration=4	1700
resolution=720p · duration=5	2100
resolution=720p · duration=6	2500
resolution=720p · duration=7	2900
resolution=720p · duration=8	3300
resolution=720p · duration=9	3700
resolution=720p · duration=10	4100
resolution=720p · duration=11	4500
resolution=720p · duration=12	4900
resolution=720p · duration=13	5300
resolution=720p · duration=14	5700
resolution=720p · duration=15	6100
resolution=1080p · duration=3	2000
resolution=1080p · duration=4	2600
resolution=1080p · duration=5	3200
resolution=1080p · duration=6	3900
resolution=1080p · duration=7	4500
resolution=1080p · duration=8	5200
resolution=1080p · duration=9	5800
resolution=1080p · duration=10	6400
resolution=1080p · duration=11	7100
resolution=1080p · duration=12	7700
resolution=1080p · duration=13	8400
resolution=1080p · duration=14	9000
resolution=1080p · duration=15	9600

happyhorse-1.0-video-edit

HappyHorse 1.0 video edit: rewrites a source video; length follows the source.

Inputs: prompt required · images optional (max 5) · video_url required

Field	Type	Values	Description
resolution	string	720p \| 1080p (default 720p)	Output resolution tier. Changes the price.
duration	integer	3–15 s (default 5)	Billing tier only — the output keeps the source length.
video_url	string	—	Source video URL (required).
images	string[]	—	Up to 5 reference images.
seed	integer	—	Random seed for reproducible output.
watermark	boolean	default false	Provider watermark (always sent; default false).
metadata.audio_setting	string	auto \| origin	Regenerate audio or keep the original track.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "happyhorse-1.0-video-edit",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "video_url": "https://example.com/source.mp4",
  "resolution": "720p",
  "duration": 5,
  "seed": 42,
  "watermark": false,
  "metadata": {
    "audio_setting": "auto"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (26)

tier	credits
resolution=720p · duration=3 · reference_video	1300
resolution=720p · duration=4 · reference_video	1700
resolution=720p · duration=5 · reference_video	2100
resolution=720p · duration=6 · reference_video	2500
resolution=720p · duration=7 · reference_video	2900
resolution=720p · duration=8 · reference_video	3300
resolution=720p · duration=9 · reference_video	3700
resolution=720p · duration=10 · reference_video	4100
resolution=720p · duration=11 · reference_video	4500
resolution=720p · duration=12 · reference_video	4900
resolution=720p · duration=13 · reference_video	5300
resolution=720p · duration=14 · reference_video	5700
resolution=720p · duration=15 · reference_video	6100
resolution=1080p · duration=3 · reference_video	2000
resolution=1080p · duration=4 · reference_video	2600
resolution=1080p · duration=5 · reference_video	3200
resolution=1080p · duration=6 · reference_video	3900
resolution=1080p · duration=7 · reference_video	4500
resolution=1080p · duration=8 · reference_video	5200
resolution=1080p · duration=9 · reference_video	5800
resolution=1080p · duration=10 · reference_video	6400
resolution=1080p · duration=11 · reference_video	7100
resolution=1080p · duration=12 · reference_video	7700
resolution=1080p · duration=13 · reference_video	8400
resolution=1080p · duration=14 · reference_video	9000
resolution=1080p · duration=15 · reference_video	9600

ByteDance · 3

seedance-2.0

ByteDance Seedance. Mode follows the media: no media = text-to-video; 1 image = first frame; 2 images = first + last frame; a video URL in images or metadata.reference_videos = reference / continuation (with-video price tier).

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	standard set plus adaptive (ratio decided by the input media)	Output aspect ratio.
resolution	string	480p \| 720p (default 480p)	Output resolution tier. Changes the price.
duration	integer	4–15 s (default 5)	Length in seconds. Changes the price.
images	string[]	—	Frame images, or video URLs (.mp4/.mov) for reference. Frames and reference media cannot be mixed.
audio	string	—	Reference audio URL; must be paired with a reference video (r2v). Max 15.2 s.
seed	integer	—	Random seed for reproducible output.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.
metadata.reference_images	string[]	≤ 9	Reference images (role reference_image).
metadata.reference_videos	string[]	≤ 3	Reference videos. Selects the with-video price tier.
metadata.reference_audios	string[]	≤ 3	Reference audio URLs.
metadata.generate_audio	boolean	—	Attach an audio track to the output.
metadata.return_last_frame	boolean	—	Also return the last frame image.
metadata.camera_fixed	boolean	—	Keep the camera fixed (stripped on seedance-2.0-fast text-to-video).
metadata.priority	integer	0–9 (default 0)	Scheduling priority; no billing effect.

seed is ignored after a provider failover.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedance-2.0",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "480p",
  "duration": 5,
  "aspect_ratio": "16:9",
  "audio": "https://example.com/reference.mp3",
  "seed": 42,
  "watermark": false,
  "metadata": {
    "reference_images": [
      "https://example.com/reference-1.jpg",
      "https://example.com/reference-2.jpg"
    ],
    "reference_videos": [
      "https://example.com/reference.mp4"
    ],
    "reference_audios": [
      "https://example.com/reference.mp3"
    ],
    "generate_audio": true,
    "return_last_frame": false,
    "camera_fixed": false,
    "priority": 0
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (48)

tier	credits
resolution=480p · duration=4	900
resolution=480p · duration=4 · reference_video	2600
resolution=480p · duration=5	1200
resolution=480p · duration=5 · reference_video	2800
resolution=480p · duration=6	1400
resolution=480p · duration=6 · reference_video	2900
resolution=480p · duration=7	1600
resolution=480p · duration=7 · reference_video	3100
resolution=480p · duration=8	1800
resolution=480p · duration=8 · reference_video	3200
resolution=480p · duration=9	2100
resolution=480p · duration=9 · reference_video	3300
resolution=480p · duration=10	2300
resolution=480p · duration=10 · reference_video	3500
resolution=480p · duration=11	2500
resolution=480p · duration=11 · reference_video	3600
resolution=480p · duration=12	2700
resolution=480p · duration=12 · reference_video	3700
resolution=480p · duration=13	3000
resolution=480p · duration=13 · reference_video	3900
resolution=480p · duration=14	3200
resolution=480p · duration=14 · reference_video	4000
resolution=480p · duration=15	3400
resolution=480p · duration=15 · reference_video	4100
resolution=720p · duration=4	2000
resolution=720p · duration=4 · reference_video	5600
resolution=720p · duration=5	2500
resolution=720p · duration=5 · reference_video	5900
resolution=720p · duration=6	2900
resolution=720p · duration=6 · reference_video	6200
resolution=720p · duration=7	3400
resolution=720p · duration=7 · reference_video	6500
resolution=720p · duration=8	3900
resolution=720p · duration=8 · reference_video	6800
resolution=720p · duration=9	4400
resolution=720p · duration=9 · reference_video	7100
resolution=720p · duration=10	4900
resolution=720p · duration=10 · reference_video	7400
resolution=720p · duration=11	5400
resolution=720p · duration=11 · reference_video	7700
resolution=720p · duration=12	5800
resolution=720p · duration=12 · reference_video	8000
resolution=720p · duration=13	6300
resolution=720p · duration=13 · reference_video	8300
resolution=720p · duration=14	6800
resolution=720p · duration=14 · reference_video	8600
resolution=720p · duration=15	7300
resolution=720p · duration=15 · reference_video	8900

seedance-2.0-fast

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	standard set plus adaptive (ratio decided by the input media)	Output aspect ratio.
resolution	string	480p \| 720p (default 480p)	Output resolution tier. Changes the price.
duration	integer	4–15 s (default 5)	Length in seconds. Changes the price.
images	string[]	—	Frame images, or video URLs (.mp4/.mov) for reference. Frames and reference media cannot be mixed.
audio	string	—	Reference audio URL; must be paired with a reference video (r2v). Max 15.2 s.
seed	integer	—	Random seed for reproducible output.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.
metadata.reference_images	string[]	≤ 9	Reference images (role reference_image).
metadata.reference_videos	string[]	≤ 3	Reference videos. Selects the with-video price tier.
metadata.reference_audios	string[]	≤ 3	Reference audio URLs.
metadata.generate_audio	boolean	—	Attach an audio track to the output.
metadata.return_last_frame	boolean	—	Also return the last frame image.
metadata.camera_fixed	boolean	—	Keep the camera fixed (stripped on seedance-2.0-fast text-to-video).
metadata.priority	integer	0–9 (default 0)	Scheduling priority; no billing effect.

seed is ignored after a provider failover.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedance-2.0-fast",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "480p",
  "duration": 5,
  "aspect_ratio": "16:9",
  "audio": "https://example.com/reference.mp3",
  "seed": 42,
  "watermark": false,
  "metadata": {
    "reference_images": [
      "https://example.com/reference-1.jpg",
      "https://example.com/reference-2.jpg"
    ],
    "reference_videos": [
      "https://example.com/reference.mp4"
    ],
    "reference_audios": [
      "https://example.com/reference.mp3"
    ],
    "generate_audio": true,
    "return_last_frame": false,
    "camera_fixed": false,
    "priority": 0
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (48)

tier	credits
resolution=480p · duration=4	800
resolution=480p · duration=4 · reference_video	2100
resolution=480p · duration=5	1000
resolution=480p · duration=5 · reference_video	2200
resolution=480p · duration=6	1100
resolution=480p · duration=6 · reference_video	2300
resolution=480p · duration=7	1300
resolution=480p · duration=7 · reference_video	2400
resolution=480p · duration=8	1500
resolution=480p · duration=8 · reference_video	2500
resolution=480p · duration=9	1700
resolution=480p · duration=9 · reference_video	2600
resolution=480p · duration=10	1900
resolution=480p · duration=10 · reference_video	2700
resolution=480p · duration=11	2000
resolution=480p · duration=11 · reference_video	2800
resolution=480p · duration=12	2200
resolution=480p · duration=12 · reference_video	2900
resolution=480p · duration=13	2400
resolution=480p · duration=13 · reference_video	3100
resolution=480p · duration=14	2600
resolution=480p · duration=14 · reference_video	3200
resolution=480p · duration=15	2800
resolution=480p · duration=15 · reference_video	3300
resolution=720p · duration=4	1600
resolution=720p · duration=4 · reference_video	4400
resolution=720p · duration=5	2000
resolution=720p · duration=5 · reference_video	4700
resolution=720p · duration=6	2400
resolution=720p · duration=6 · reference_video	4900
resolution=720p · duration=7	2800
resolution=720p · duration=7 · reference_video	5100
resolution=720p · duration=8	3200
resolution=720p · duration=8 · reference_video	5400
resolution=720p · duration=9	3500
resolution=720p · duration=9 · reference_video	5600
resolution=720p · duration=10	3900
resolution=720p · duration=10 · reference_video	5800
resolution=720p · duration=11	4300
resolution=720p · duration=11 · reference_video	6000
resolution=720p · duration=12	4700
resolution=720p · duration=12 · reference_video	6300
resolution=720p · duration=13	5100
resolution=720p · duration=13 · reference_video	6500
resolution=720p · duration=14	5500
resolution=720p · duration=14 · reference_video	6700
resolution=720p · duration=15	5900
resolution=720p · duration=15 · reference_video	7000

seedance-2.5

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
aspect_ratio	string	standard set plus adaptive (ratio decided by the input media); rejected on first/last-frame tasks	Output aspect ratio.
resolution	string	480p \| 720p (default 480p) — (1080p costs 2.25× 720p)	Output resolution tier. Changes the price.
duration	integer	4–30 s (default 5)	Length in seconds. Changes the price.
images	string[]	—	Frame images, or video URLs (.mp4/.mov) for reference. Frames and reference media cannot be mixed.
audio	string	—	Reference audio URL.
seed	integer	—	Random seed for reproducible output.
watermark	boolean	—	Provider watermark. Only forwarded when you send it; omitted follows the provider default.
metadata.reference_images	string[]	≤ 30	Reference images (role reference_image).
metadata.reference_videos	string[]	≤ 10	Reference videos. Selects the with-video price tier.
metadata.reference_audios	string[]	≤ 10	Reference audio URLs.
metadata.generate_audio	boolean	—	Attach an audio track to the output.
metadata.return_last_frame	boolean	—	Also return the last frame image.

Send aspect_ratio "adaptive" whenever the request carries a reference video.
Edit-style prompts ("replace / delete…" over a reference video) are not served on this line; use reference semantics.
seed is ignored after a provider failover.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedance-2.5",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "480p",
  "duration": 5,
  "aspect_ratio": "16:9",
  "audio": "https://example.com/reference.mp3",
  "seed": 42,
  "watermark": false,
  "metadata": {
    "reference_images": [
      "https://example.com/reference-1.jpg",
      "https://example.com/reference-2.jpg"
    ],
    "reference_videos": [
      "https://example.com/reference.mp4"
    ],
    "reference_audios": [
      "https://example.com/reference.mp3"
    ],
    "generate_audio": true,
    "return_last_frame": false
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (108)

tier	credits
resolution=480p · duration=4	1300
resolution=480p · duration=4 · reference_video	6400
resolution=480p · duration=5	1600
resolution=480p · duration=5 · reference_video	6600
resolution=480p · duration=6	1900
resolution=480p · duration=6 · reference_video	6800
resolution=480p · duration=7	2200
resolution=480p · duration=7 · reference_video	7000
resolution=480p · duration=8	2500
resolution=480p · duration=8 · reference_video	7100
resolution=480p · duration=9	2800
resolution=480p · duration=9 · reference_video	7300
resolution=480p · duration=10	3100
resolution=480p · duration=10 · reference_video	7500
resolution=480p · duration=11	3400
resolution=480p · duration=11 · reference_video	7700
resolution=480p · duration=12	3800
resolution=480p · duration=12 · reference_video	7900
resolution=480p · duration=13	4100
resolution=480p · duration=13 · reference_video	8100
resolution=480p · duration=14	4400
resolution=480p · duration=14 · reference_video	8300
resolution=480p · duration=15	4700
resolution=480p · duration=15 · reference_video	8500
resolution=480p · duration=16	5000
resolution=480p · duration=16 · reference_video	8600
resolution=480p · duration=17	5300
resolution=480p · duration=17 · reference_video	8800
resolution=480p · duration=18	5600
resolution=480p · duration=18 · reference_video	9000
resolution=480p · duration=19	6000
resolution=480p · duration=19 · reference_video	9200
resolution=480p · duration=20	6300
resolution=480p · duration=20 · reference_video	9400
resolution=480p · duration=21	6600
resolution=480p · duration=21 · reference_video	9600
resolution=480p · duration=22	6900
resolution=480p · duration=22 · reference_video	9800
resolution=480p · duration=23	7200
resolution=480p · duration=23 · reference_video	10000
resolution=480p · duration=24	7500
resolution=480p · duration=24 · reference_video	10100
resolution=480p · duration=25	7800
resolution=480p · duration=25 · reference_video	10300
resolution=480p · duration=26	8100
resolution=480p · duration=26 · reference_video	10500
resolution=480p · duration=27	8500
resolution=480p · duration=27 · reference_video	10700
resolution=480p · duration=28	8800
resolution=480p · duration=28 · reference_video	10900
resolution=480p · duration=29	9100
resolution=480p · duration=29 · reference_video	11100
resolution=480p · duration=30	9400
resolution=480p · duration=30 · reference_video	11300
resolution=720p · duration=4	2800
resolution=720p · duration=4 · reference_video	14400
resolution=720p · duration=5	3500
resolution=720p · duration=5 · reference_video	14800
resolution=720p · duration=6	4200
resolution=720p · duration=6 · reference_video	15200
resolution=720p · duration=7	4900
resolution=720p · duration=7 · reference_video	15600
resolution=720p · duration=8	5600
resolution=720p · duration=8 · reference_video	16100
resolution=720p · duration=9	6300
resolution=720p · duration=9 · reference_video	16500
resolution=720p · duration=10	7000
resolution=720p · duration=10 · reference_video	16900
resolution=720p · duration=11	7700
resolution=720p · duration=11 · reference_video	17300
resolution=720p · duration=12	8500
resolution=720p · duration=12 · reference_video	17700
resolution=720p · duration=13	9200
resolution=720p · duration=13 · reference_video	18200
resolution=720p · duration=14	9900
resolution=720p · duration=14 · reference_video	18600
resolution=720p · duration=15	10600
resolution=720p · duration=15 · reference_video	19000
resolution=720p · duration=16	11300
resolution=720p · duration=16 · reference_video	19400
resolution=720p · duration=17	12000
resolution=720p · duration=17 · reference_video	19900
resolution=720p · duration=18	12700
resolution=720p · duration=18 · reference_video	20300
resolution=720p · duration=19	13400
resolution=720p · duration=19 · reference_video	20700
resolution=720p · duration=20	14100
resolution=720p · duration=20 · reference_video	21100
resolution=720p · duration=21	14800
resolution=720p · duration=21 · reference_video	21500
resolution=720p · duration=22	15500
resolution=720p · duration=22 · reference_video	22000
resolution=720p · duration=23	16200
resolution=720p · duration=23 · reference_video	22400
resolution=720p · duration=24	16900
resolution=720p · duration=24 · reference_video	22800
resolution=720p · duration=25	17600
resolution=720p · duration=25 · reference_video	23200
resolution=720p · duration=26	18300
resolution=720p · duration=26 · reference_video	23700
resolution=720p · duration=27	19000
resolution=720p · duration=27 · reference_video	24100
resolution=720p · duration=28	19700
resolution=720p · duration=28 · reference_video	24500
resolution=720p · duration=29	20400
resolution=720p · duration=29 · reference_video	24900
resolution=720p · duration=30	21100
resolution=720p · duration=30 · reference_video	25400

MiniMax · 1

minimax-h3

MiniMax Hailuo H3. Text-to-video (aspect_ratio required, not adaptive), image-to-video (1 image = first frame, 2 = first + last), or reference-to-video (metadata.reference_*). Frames and reference media cannot be mixed.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
resolution	string	768P \| 2K (default 768P)	Output resolution tier. Changes the price.
duration	integer	4–15 s (default 5)	Length in seconds. Changes the price.
aspect_ratio	string	standard set plus adaptive (not for text-to-video); default 16:9 when you send nothing	Output aspect ratio.
images	string[]	—	Up to 2 frame images.
metadata.reference_images	string[]	≤ 9 (the first 5 are free; more add a small per-image charge)	Reference images.
metadata.reference_videos	string[]	≤ 3, ≤ 15 s total	Reference videos — with-video price tier, settled to the real input length on completion.
metadata.reference_audios	string[]	≤ 3	Reference audio (must be paired with an image or video on the primary line).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "768P",
  "duration": 5,
  "aspect_ratio": "16:9",
  "metadata": {
    "reference_images": [
      "https://example.com/reference-1.jpg",
      "https://example.com/reference-2.jpg"
    ],
    "reference_videos": [
      "https://example.com/reference.mp4"
    ],
    "reference_audios": [
      "https://example.com/reference.mp3"
    ]
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (48)

tier	credits
resolution=768P · duration=4	1100
resolution=768P · duration=4 · reference_video	5100
resolution=768P · duration=5	1300
resolution=768P · duration=5 · reference_video	5300
resolution=768P · duration=6	1600
resolution=768P · duration=6 · reference_video	5600
resolution=768P · duration=7	1900
resolution=768P · duration=7 · reference_video	5900
resolution=768P · duration=8	2100
resolution=768P · duration=8 · reference_video	6100
resolution=768P · duration=9	2400
resolution=768P · duration=9 · reference_video	6400
resolution=768P · duration=10	2700
resolution=768P · duration=10 · reference_video	6700
resolution=768P · duration=11	2900
resolution=768P · duration=11 · reference_video	6900
resolution=768P · duration=12	3200
resolution=768P · duration=12 · reference_video	7200
resolution=768P · duration=13	3500
resolution=768P · duration=13 · reference_video	7500
resolution=768P · duration=14	3700
resolution=768P · duration=14 · reference_video	7700
resolution=768P · duration=15	4000
resolution=768P · duration=15 · reference_video	8000
resolution=2K · duration=4	1700
resolution=2K · duration=4 · reference_video	8200
resolution=2K · duration=5	2200
resolution=2K · duration=5 · reference_video	8700
resolution=2K · duration=6	2600
resolution=2K · duration=6 · reference_video	9100
resolution=2K · duration=7	3000
resolution=2K · duration=7 · reference_video	9500
resolution=2K · duration=8	3500
resolution=2K · duration=8 · reference_video	10000
resolution=2K · duration=9	3900
resolution=2K · duration=9 · reference_video	10400
resolution=2K · duration=10	4300
resolution=2K · duration=10 · reference_video	10800
resolution=2K · duration=11	4800
resolution=2K · duration=11 · reference_video	11300
resolution=2K · duration=12	5200
resolution=2K · duration=12 · reference_video	11700
resolution=2K · duration=13	5600
resolution=2K · duration=13 · reference_video	12100
resolution=2K · duration=14	6100
resolution=2K · duration=14 · reference_video	12600
resolution=2K · duration=15	6500
resolution=2K · duration=15 · reference_video	13000

Black Forest Labs · 1

flux-3-video

Black Forest Labs FLUX 3 Video. Mode follows the assets: no media = text-to-video; images = keyframes (1 start, 2 start + end, 3–10 spaced evenly, order is semantic); a video URL = continuation (~2.4× the price). Keyframes and a continuation video cannot be combined.

Inputs: prompt required · images optional (max 4)

Field	Type	Values	Description
resolution	string	hd \| fhd (default hd)	Output resolution tier. Changes the price.
duration	integer	5–20 s (default 5)	Length in seconds. Changes the price.
aspect_ratio	string	21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16 (omit for auto)	Output aspect ratio.
images	string[]	—	1–4 keyframe images.
reference_videos	string[]	exactly 1	Continuation source video.
draft	boolean	—	Low-quality preview at ~1/3 price; hd only (draft + fhd is a 400).
metadata.generate_audio	boolean	default true	Audio track on/off (same price).

Continuation is billed for the requested seconds even if the provider returns a shorter clip.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "flux-3-video",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "hd",
  "duration": 5,
  "aspect_ratio": "16:9",
  "reference_videos": [
    "https://example.com/reference.mp4"
  ],
  "draft": false,
  "metadata": {
    "generate_audio": true
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (96)

tier	credits
resolution=hd · duration=5	2300
resolution=hd · duration=5 · draft	800
resolution=hd · duration=5 · reference_video	5500
resolution=hd · duration=5 · reference_video · draft	1600
resolution=hd · duration=6	2700
resolution=hd · duration=6 · draft	1000
resolution=hd · duration=6 · reference_video	6600
resolution=hd · duration=6 · reference_video · draft	1900
resolution=hd · duration=7	3200
resolution=hd · duration=7 · draft	1100
resolution=hd · duration=7 · reference_video	7600
resolution=hd · duration=7 · reference_video · draft	2200
resolution=hd · duration=8	3600
resolution=hd · duration=8 · draft	1300
resolution=hd · duration=8 · reference_video	8700
resolution=hd · duration=8 · reference_video · draft	2600
resolution=hd · duration=9	4100
resolution=hd · duration=9 · draft	1400
resolution=hd · duration=9 · reference_video	9800
resolution=hd · duration=9 · reference_video · draft	2900
resolution=hd · duration=10	4500
resolution=hd · duration=10 · draft	1600
resolution=hd · duration=10 · reference_video	10900
resolution=hd · duration=10 · reference_video · draft	3200
resolution=hd · duration=11	5000
resolution=hd · duration=11 · draft	1800
resolution=hd · duration=11 · reference_video	12000
resolution=hd · duration=11 · reference_video · draft	3500
resolution=hd · duration=12	5400
resolution=hd · duration=12 · draft	1900
resolution=hd · duration=12 · reference_video	13100
resolution=hd · duration=12 · reference_video · draft	3800
resolution=hd · duration=13	5900
resolution=hd · duration=13 · draft	2100
resolution=hd · duration=13 · reference_video	14200
resolution=hd · duration=13 · reference_video · draft	4200
resolution=hd · duration=14	6300
resolution=hd · duration=14 · draft	2200
resolution=hd · duration=14 · reference_video	15300
resolution=hd · duration=14 · reference_video · draft	4500
resolution=hd · duration=15	6800
resolution=hd · duration=15 · draft	2400
resolution=hd · duration=15 · reference_video	16400
resolution=hd · duration=15 · reference_video · draft	4800
resolution=hd · duration=16	7200
resolution=hd · duration=16 · draft	2600
resolution=hd · duration=16 · reference_video	17500
resolution=hd · duration=16 · reference_video · draft	5100
resolution=hd · duration=17	7700
resolution=hd · duration=17 · draft	2700
resolution=hd · duration=17 · reference_video	18600
resolution=hd · duration=17 · reference_video · draft	5400
resolution=hd · duration=18	8200
resolution=hd · duration=18 · draft	2900
resolution=hd · duration=18 · reference_video	19700
resolution=hd · duration=18 · reference_video · draft	5800
resolution=hd · duration=19	8600
resolution=hd · duration=19 · draft	3000
resolution=hd · duration=19 · reference_video	20800
resolution=hd · duration=19 · reference_video · draft	6100
resolution=hd · duration=20	9100
resolution=hd · duration=20 · draft	3200
resolution=hd · duration=20 · reference_video	21800
resolution=hd · duration=20 · reference_video · draft	6400
resolution=fhd · duration=5	3900
resolution=fhd · duration=5 · reference_video	7100
resolution=fhd · duration=6	4600
resolution=fhd · duration=6 · reference_video	8500
resolution=fhd · duration=7	5400
resolution=fhd · duration=7 · reference_video	9900
resolution=fhd · duration=8	6200
resolution=fhd · duration=8 · reference_video	11300
resolution=fhd · duration=9	7000
resolution=fhd · duration=9 · reference_video	12700
resolution=fhd · duration=10	7700
resolution=fhd · duration=10 · reference_video	14100
resolution=fhd · duration=11	8500
resolution=fhd · duration=11 · reference_video	15500
resolution=fhd · duration=12	9300
resolution=fhd · duration=12 · reference_video	16900
resolution=fhd · duration=13	10000
resolution=fhd · duration=13 · reference_video	18400
resolution=fhd · duration=14	10800
resolution=fhd · duration=14 · reference_video	19800
resolution=fhd · duration=15	11600
resolution=fhd · duration=15 · reference_video	21200
resolution=fhd · duration=16	12400
resolution=fhd · duration=16 · reference_video	22600
resolution=fhd · duration=17	13100
resolution=fhd · duration=17 · reference_video	24000
resolution=fhd · duration=18	13900
resolution=fhd · duration=18 · reference_video	25400
resolution=fhd · duration=19	14700
resolution=fhd · duration=19 · reference_video	26800
resolution=fhd · duration=20	15500
resolution=fhd · duration=20 · reference_video	28200

Agnes · 3

agnes-video-2.5

Agnes Video 2.5. Modes text / keyframe / reference are inferred from the material (1 image = first frame, 2 = first + last, 3+ = reference; any reference_* list or audio = reference) or set explicitly with agnes_mode.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
resolution	string	720P \| 960P \| 2K (default 720P)	Output resolution tier. Changes the price.
duration	integer	4–12 s (default 5) — 4–12 (default 5)	Length in seconds. Changes the price.
aspect_ratio	string	21:9, 16:9, 4:3, 1:1, 3:4, 9:16 (default 16:9; anything else is a 400)	Output aspect ratio.
images	string[]	—	Up to 10 images (keyframes or references).
metadata.agnes_mode	string	text \| keyframe \| reference	Force a mode; a mode/material mismatch is a 400 before any charge.
metadata.first_frame	string	—	Explicit first-frame slot (keyframe mode).
metadata.last_frame	string	—	Explicit last-frame slot — the only way to send a last frame alone.
metadata.reference_images	string[]	beyond 5 images each adds a small per-image charge	Reference images (<Picture N> in the prompt).
metadata.reference_videos	(string \| {url, start_seconds, require_audio})[]	—	Reference videos (<Video N>). Each reference video is billed as one extra output-length of input seconds.
metadata.reference_audios	string[]	—	Reference audio (<Audio N>).
seed	integer	—	Random seed for reproducible output.

Billed per second of output (plus estimated reference-video seconds). negative_prompt / watermark are not supported.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "agnes-video-2.5",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "720P",
  "duration": 5,
  "aspect_ratio": "16:9",
  "seed": 42,
  "metadata": {
    "agnes_mode": "keyframe",
    "first_frame": "https://example.com/first-frame.jpg",
    "last_frame": "https://example.com/last-frame.jpg",
    "reference_images": [
      "https://example.com/reference-1.jpg",
      "https://example.com/reference-2.jpg"
    ],
    "reference_videos": [
      "https://example.com/reference.mp4"
    ],
    "reference_audios": [
      "https://example.com/reference.mp3"
    ]
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (54)

tier	credits
resolution=720P · duration=4	330
resolution=720P · duration=4 · reference_video	670
resolution=720P · duration=5	420
resolution=720P · duration=5 · reference_video	830
resolution=720P · duration=6	500
resolution=720P · duration=6 · reference_video	1000
resolution=720P · duration=7	580
resolution=720P · duration=7 · reference_video	1170
resolution=720P · duration=8	670
resolution=720P · duration=8 · reference_video	1330
resolution=720P · duration=9	750
resolution=720P · duration=9 · reference_video	1500
resolution=720P · duration=10	830
resolution=720P · duration=10 · reference_video	1670
resolution=720P · duration=11	920
resolution=720P · duration=11 · reference_video	1830
resolution=720P · duration=12	1000
resolution=720P · duration=12 · reference_video	2000
resolution=960P · duration=4	530
resolution=960P · duration=4 · reference_video	1070
resolution=960P · duration=5	670
resolution=960P · duration=5 · reference_video	1330
resolution=960P · duration=6	800
resolution=960P · duration=6 · reference_video	1600
resolution=960P · duration=7	930
resolution=960P · duration=7 · reference_video	1860
resolution=960P · duration=8	1070
resolution=960P · duration=8 · reference_video	2130
resolution=960P · duration=9	1200
resolution=960P · duration=9 · reference_video	2400
resolution=960P · duration=10	1330
resolution=960P · duration=10 · reference_video	2660
resolution=960P · duration=11	1470
resolution=960P · duration=11 · reference_video	2930
resolution=960P · duration=12	1600
resolution=960P · duration=12 · reference_video	3200
resolution=2K · duration=4	730
resolution=2K · duration=4 · reference_video	1470
resolution=2K · duration=5	920
resolution=2K · duration=5 · reference_video	1830
resolution=2K · duration=6	1100
resolution=2K · duration=6 · reference_video	2200
resolution=2K · duration=7	1280
resolution=2K · duration=7 · reference_video	2560
resolution=2K · duration=8	1470
resolution=2K · duration=8 · reference_video	2930
resolution=2K · duration=9	1650
resolution=2K · duration=9 · reference_video	3300
resolution=2K · duration=10	1830
resolution=2K · duration=10 · reference_video	3660
resolution=2K · duration=11	2020
resolution=2K · duration=11 · reference_video	4030
resolution=2K · duration=12	2200
resolution=2K · duration=12 · reference_video	4400

agnes-video-2.5-flash

Agnes Video 2.5 Flash — same contract as agnes-video-2.5, 720P only, at most 5 reference images, no reference videos.

Inputs: prompt required · images optional (max 5)

Field	Type	Values	Description
resolution	string	720P (default 720P)	Output resolution tier. Changes the price.
duration	integer	4–12 s (default 5) — 4–12 (default 5)	Length in seconds. Changes the price.
aspect_ratio	string	21:9, 16:9, 4:3, 1:1, 3:4, 9:16 (default 16:9)	Output aspect ratio.
images	string[]	—	Up to 5 images.
metadata.agnes_mode	string	text \| keyframe \| reference	Force a mode.
metadata.first_frame	string	—	Explicit first-frame slot.
metadata.last_frame	string	—	Explicit last-frame slot.
metadata.reference_images	string[]	≤ 5	Reference images.
metadata.reference_audios	string[]	—	Reference audio.
seed	integer	—	Random seed for reproducible output.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "agnes-video-2.5-flash",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "720P",
  "duration": 5,
  "aspect_ratio": "16:9",
  "seed": 42,
  "metadata": {
    "agnes_mode": "keyframe",
    "first_frame": "https://example.com/first-frame.jpg",
    "last_frame": "https://example.com/last-frame.jpg",
    "reference_images": [
      "https://example.com/reference-1.jpg",
      "https://example.com/reference-2.jpg"
    ],
    "reference_audios": [
      "https://example.com/reference.mp3"
    ]
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (9)

tier	credits
resolution=720P · duration=4	330
resolution=720P · duration=5	420
resolution=720P · duration=6	500
resolution=720P · duration=7	580
resolution=720P · duration=8	670
resolution=720P · duration=9	750
resolution=720P · duration=10	830
resolution=720P · duration=11	920
resolution=720P · duration=12	1000

agnes-video-v2.0

Agnes Video 2.0. 1 image = first frame; 2+ images = keyframe transition (default) or multi-reference with agnes_mode = reference.

Inputs: prompt required · images optional (max 10)

Field	Type	Values	Description
resolution	string	720p \| 480p \| 1080p (default 720p)	Output resolution tier. Changes the price.
duration	integer	3 \| 5 \| 10 \| 18 s (default 5) — 3 \| 5 \| 10 \| 18 (other values round up to the next tier)	Length in seconds. Changes the price.
aspect_ratio	string	16:9, 9:16, 1:1, 4:3, 3:4 (others snap; default 16:9)	Output aspect ratio.
images	string[]	—	Up to 10 images.
metadata.agnes_mode	string	keyframes (default) \| reference	How 2+ images are used.
seed	integer	—	Random seed for reproducible output.
metadata.negative_prompt	string	—	What to avoid.

Billed per second at the rounded-up duration tier.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "agnes-video-v2.0",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "resolution": "720p",
  "duration": 5,
  "aspect_ratio": "16:9",
  "seed": 42,
  "metadata": {
    "agnes_mode": "keyframe",
    "negative_prompt": "blurry, low quality, text"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (12)

tier	credits
resolution=720p · duration=3	45
resolution=720p · duration=5	75
resolution=720p · duration=10	150
resolution=720p · duration=18	270
resolution=480p · duration=3	45
resolution=480p · duration=5	75
resolution=480p · duration=10	150
resolution=480p · duration=18	270
resolution=1080p · duration=3	45
resolution=1080p · duration=5	75
resolution=1080p · duration=10	150
resolution=1080p · duration=18	270

xAI · 2

grok-imagine-video

xAI Grok Imagine — text-to-video or image-to-video.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
resolution	string	480p \| 720p \| 1080p	Output resolution tier. Changes the price.
duration	integer	1–15 s (default 5)	Length in seconds. Billed per second.
images	string[]	—	Optional: 1 image.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "grok-imagine-video",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "duration": 5,
  "aspect_ratio": "16:9"
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (15)

tier	credits
duration=1	100
duration=2	200
duration=3	300
duration=4	400
duration=5	500
duration=6	600
duration=7	700
duration=8	800
duration=9	900
duration=10	1000
duration=11	1100
duration=12	1200
duration=13	1300
duration=14	1400
duration=15	1500

grok-imagine-video-1.5

xAI Grok Imagine 1.5 — image-to-video only.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
resolution	string	480p \| 720p \| 1080p	Output resolution tier. Changes the price.
duration	integer	1–15 s (default 5)	Length in seconds. Billed per second.
images	string[]	—	Required: 1 image.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "grok-imagine-video-1.5",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "duration": 5,
  "aspect_ratio": "16:9"
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (15)

tier	credits
duration=1	150
duration=2	300
duration=3	450
duration=4	600
duration=5	750
duration=6	900
duration=7	1050
duration=8	1200
duration=9	1350
duration=10	1500
duration=11	1650
duration=12	1800
duration=13	1950
duration=14	2100
duration=15	2250

OpenAI · 1

sora-2

OpenAI Sora 2 text-to-video / image-to-video.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
duration	integer	4 \| 8 \| 12 s (default 4) — 4 \| 8 \| 12 (others snap to the nearest)	Length in seconds. Changes the price.
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
images	string[]	—	Optional: 1 image.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "sora-2",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "duration": 4,
  "aspect_ratio": "16:9"
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (3)

tier	credits
duration=4	600
duration=8	1200
duration=12	1800

Google · 1

veo3.1

Google Veo 3.1 (fast). Fixed 8-second clips; an input image selects image-to-video.

Inputs: prompt required · images optional (max 3)

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
resolution	string	720p \| 1080p	Output resolution tier. Changes the price.
images	string[]	—	Up to 3 images (the first is the start frame).

duration is fixed at 8 s and cannot be changed.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "veo3.1",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg",
    "https://example.com/input-2.jpg"
  ],
  "aspect_ratio": "16:9"
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (2)

tier	credits
flat	1500
image_to_video	1500

Runway · 1

runway-gen4-turbo

Runway Gen-4 Turbo — image-to-video only.

Inputs: prompt required · 1 image required (max 1)

Field	Type	Values	Description
images	string[]	—	Required: 1 image.
duration	integer	5 \| 10 s (default 5)	Length in seconds. Changes the price.
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "runway-gen4-turbo",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "duration": 5,
  "aspect_ratio": "16:9"
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Price tiers (2)

tier	credits
duration=5	200
duration=10	400

Luma · 1

luma-ray-2-flash

Luma Ray 2 Flash text-to-video.

Inputs: prompt required

Field	Type	Values	Description
aspect_ratio	string	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21 (unsupported values snap to the nearest)	Output aspect ratio.
metadata.loop	boolean	—	Seamless loop.

Flat price.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "luma-ray-2-flash",
  "prompt": "A small red boat drifting on a calm lake at sunrise, gentle waves",
  "aspect_ratio": "16:9",
  "metadata": {
    "loop": false
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Audio models

Text-to-speech models read prompt as the text to speak (per 1,000 characters); music models read it as a description (per task or per second). Per-second models reserve a maximum at submit and settle on the real length.

ElevenLabs · 3

eleven-v3

ElevenLabs Eleven v3 text-to-speech with expressive audio tags like [whispering], [laughing].

Inputs: prompt required

Field	Type	Values	Description
metadata.voice_id	string	default: George (JBFqnCBsd6RMkjVDRZzb)	ElevenLabs voice id.
metadata.stability	number	0–1	Lower = wider emotional range.
metadata.similarity_boost	number	0–1	How closely to match the original voice.
metadata.style	number	—	Style exaggeration (non-zero adds latency).
metadata.speed	number	1.0 = normal	Speaking speed.
metadata.use_speaker_boost	boolean	—	Boost similarity to the original speaker.
metadata.language_code	string	ISO 639-1	Omit for auto-detection.

prompt is the text to speak, max 5,000 characters. Billed per 1,000 characters at submit. Output MP3.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "eleven-v3",
  "prompt": "Hello from GlobalGPT. This is a short test sentence.",
  "metadata": {
    "voice_id": "JBFqnCBsd6RMkjVDRZzb",
    "stability": 0.5,
    "similarity_boost": 0.75,
    "style": 0.2,
    "speed": 1,
    "use_speaker_boost": true,
    "language_code": "en"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

eleven-multilingual-v2

ElevenLabs Multilingual v2 text-to-speech — 29 languages, steady long-form narration (audio tags are read literally).

Inputs: prompt required

Field	Type	Values	Description
metadata.voice_id	string	default: George (JBFqnCBsd6RMkjVDRZzb)	ElevenLabs voice id.
metadata.stability	number	0–1	Lower = wider emotional range.
metadata.similarity_boost	number	0–1	How closely to match the original voice.
metadata.style	number	—	Style exaggeration (non-zero adds latency).
metadata.speed	number	1.0 = normal	Speaking speed.
metadata.use_speaker_boost	boolean	—	Boost similarity to the original speaker.
metadata.language_code	string	ISO 639-1	Omit for auto-detection.

prompt is the text to speak, max 10,000 characters. Billed per 1,000 characters at submit. Output MP3.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "eleven-multilingual-v2",
  "prompt": "Hello from GlobalGPT. This is a short test sentence.",
  "metadata": {
    "voice_id": "JBFqnCBsd6RMkjVDRZzb",
    "stability": 0.5,
    "similarity_boost": 0.75,
    "style": 0.2,
    "speed": 1,
    "use_speaker_boost": true,
    "language_code": "en"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

eleven-music-v2

ElevenLabs Music v2 text-to-music. prompt describes genre, mood, instruments, vocals, structure.

Inputs: prompt required

Field	Type	Values	Description
duration	integer	10–300 s (default 30) — 10–300 (default 30)	Requested length in seconds; billed per second at submit.
metadata.force_instrumental	boolean	—	No vocals.

Output MP3 (44.1 kHz, 128 kbps).

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "eleven-music-v2",
  "prompt": "A calm piano melody with soft strings, 90 BPM",
  "duration": 30,
  "metadata": {
    "force_instrumental": false
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Alibaba · 1

qwen-audio-3.0-tts-flash

Alibaba Qwen-Audio TTS — Chinese-first multilingual speech, 49 voices including Chinese dialects. Output WAV.

Inputs: prompt required

Field	Type	Values	Description
metadata.voice	string	e.g. Cherry (default), Ethan, Jada, Dylan	Voice name; an unknown voice fails the task.
metadata.language_type	string	Auto (default), Chinese, English, …	Language hint; set it when using dialect voices.

prompt is the text to speak, max 4,096 characters. Billed per 1,000 characters.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "qwen-audio-3.0-tts-flash",
  "prompt": "Hello from GlobalGPT. This is a short test sentence.",
  "metadata": {
    "voice": "Cherry",
    "language_type": "Auto"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

ByteDance · 1

seed-audio-1.0

ByteDance Seed-Audio 1.0 — finished audio from one prompt: multi-character dialogue, emotion, music and effects (up to 120 s). Exactly one reference source: audio, images or metadata.speaker.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
audio	string	—	Reference audio URL (≤ 30 s, ≤ 10 MB; wav/mp3/pcm/ogg).
images	string[]	—	Optional: 1 reference image (jpeg/png/webp, ≤ 10 MB).
metadata.speaker	string	—	Doubao TTS voice id or cloned voice id.
metadata.speech_rate	number	—	Speaking rate.
metadata.loudness_rate	number	—	Loudness.
metadata.pitch_rate	number	—	Pitch.

prompt max 3,000 characters; address a reference audio as @音频1.
Reserved for 120 s at submit, settled to the real output length.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seed-audio-1.0",
  "prompt": "Hello from GlobalGPT. This is a short test sentence.",
  "images": [
    "https://example.com/input-1.jpg"
  ],
  "audio": "https://example.com/reference.mp3",
  "metadata": {
    "speaker": "zh_female_xiaohe",
    "speech_rate": 1,
    "loudness_rate": 1,
    "pitch_rate": 1
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Mureka · 2

mureka-v9-song

Mureka V9 — sings your lyrics as a full song. prompt = the lyrics (section tags like [Verse] / [Chorus] supported).

Inputs: prompt required

Field	Type	Values	Description
metadata.style_prompt	string	≤ 1,024 characters	Genre, mood, instruments, vocals.

prompt max 3,000 characters. Flat price per track.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "mureka-v9-song",
  "prompt": "[Verse]\nCity lights are fading out\n[Chorus]\nWe keep on driving",
  "metadata": {
    "style_prompt": "warm acoustic folk, female vocal"
  }
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

mureka-v9-bgm

Mureka V9 — instrumental background music from a description.

Inputs: prompt required

No optional parameters — send model and prompt.

prompt max 1,024 characters. Flat price per track.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "mureka-v9-bgm",
  "prompt": "A calm piano melody with soft strings, 90 BPM"
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Google · 2

lyria-3-clip

Google Lyria 3 — 30-second music clip from a prompt (genre, mood, instruments, BPM). English prompts recommended.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
images	string[]	—	Optional: 1 reference image whose mood inspires the music.

Flat price. seed / negative_prompt are ignored.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "lyria-3-clip",
  "prompt": "A calm piano melody with soft strings, 90 BPM",
  "images": [
    "https://example.com/input-1.jpg"
  ]
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

lyria-3-pro

Google Lyria 3 Pro — music up to 3 minutes, length steered by the prompt.

Inputs: prompt required · images optional (max 1)

Field	Type	Values	Description
images	string[]	—	Optional: 1 reference image whose mood inspires the music.

Flat price. seed / negative_prompt are ignored.

Example request (all parameters)

# 1) submit
curl https://api2.glbgpt.com/ai-api/open/v1/tasks \
  -H "Authorization: Bearer sk-glb-..." \
  -H "Content-Type: application/json" \
  -d '{
  "model": "lyria-3-pro",
  "prompt": "A calm piano melody with soft strings, 90 BPM",
  "images": [
    "https://example.com/input-1.jpg"
  ]
}'

# 2) poll until status is "succeeded" or "failed" (every 5-10 s)
curl https://api2.glbgpt.com/ai-api/open/v1/tasks/TASK_ID -H "Authorization: Bearer sk-glb-..."

Usage

GET /v1/usage?start_date=YYYY-MM-DD&end_date=YYYY-MM-DD (UTC, up to 90 days, default last 30) returns per-day totals and a per-model breakdown of credits, requests and tokens for API traffic only. The same numbers are on the Usage page.

Errors & limits

Errors use the OpenAI envelope: {"error": {"message", "type", "code", "trace_id"}} with a real HTTP status. 429 responses include Retry-After.

HTTP	code	Meaning
400	invalid_request	Body validation failed — the message says which field.
400	upstream_rejected / context_length_exceeded / content_policy_violation	The model provider refused the chat request.
401	missing_api_key / invalid_api_key	Send Authorization: Bearer sk-glb-…
402	insufficient_credit	Not enough paid credits. Top up or subscribe.
402	paid_only_channel	The API never spends free or trial credits.
402	plan_required	The model is not included in your plan.
404	not_found / model_not_found	Unknown model, retired model (message names the successor), or task not yours.
409	idempotency_in_progress	Same Idempotency-Key is still being processed.
422	idempotency_key_reused	Same Idempotency-Key with a different body.
429	rate_limit_exceeded / too_many_auth_failures	Per-account limits (60 req/min, 5 chat streams, 10 media tasks in flight) or repeated bad keys from one IP. Honour Retry-After.
502	upstream_error / model_unavailable	Provider failed or has no capacity for this model right now. Retry or switch model.

Rate limits per account: 60 requests per minute, 5 concurrent chat streams, 10 media tasks in flight. Ask us if you need more.