Seedance 2.0 is quickly becoming one of the most powerful AI video models for creators who want true cinematic results. With strong frame-to-frame consistency, realistic physics, native audio generation, and advanced camera control, Seedance 2.0 allows you to create film-quality AI videos using well-structured prompts and reference materials.
在本指南中，您将学习如何编写高性能的 Seedance 2.0 prompts, 您还可以按照经过验证的工作流程，使用 5 个真实的复制粘贴提示示例来创建病毒式人工智能视频。.
如果您想知道 more comprehensive and advanced usage methods of Seedance, you can refer to Seedance 2.0 终极提示指南.
GlobalGPT has officially integrated Seedance 2.0, removing all phone verification and regional restrictions. For $10.8/month on the Pro Plan, you get instant access to execute Seedance 2.0 prompts without waitlists. You can switch between Seedance 2.0 and other cinematic models like Veo 3.1 or Kling in a single workspace.
为什么说 Seedance 2.0 会改变人工智能视频提示的游戏规则？
Seedance 2.0 的突出之处在于它能理解 “导演级别 ”的提示，而不仅仅是简单的文本到视频。在 Seedance 2.0 评论, 该模型处理复杂指令的能力经常被认为是其最强大的资产。.
There are 9 key features that creators call a category killer:
- Near-perfect character, clothing, and lighting consistency across frames
- 强大的参考系统：
- 一张图片锁定视觉风格
- 一个视频锁定摄像机的移动和动作
- 一个音轨，锁定节奏和步调
- 根据单一描述自动拍摄多张照片讲故事
- Native audio + video generation with synced dialogue and sound effects (note: consider privacy risks)
- 布料、液体、重力和战斗都具有很强的物理真实感
这意味着您的 Seedance 2.0 提示不再只是文本。它变成了完整的生产指令。.
什么是高质量的 Seedance 2.0 提示？
Seedance 2.0 的最佳提示遵循以下原则 导演式结构:
主题 + 动作 + 镜头 + 场景 + 风格 + 限制条件
您的要求越具体，结果就越具有电影感和连贯性。.
使用 @Image1 作为主要主题参考。.
Create a video where [subject] [action] in [scene].
The camera [camera movement].
The visual style is [style, lighting, mood].
Keep [face/product/logo/clothing] consistent throughout the video.
Avoid distortion, flickering, extra objects, unreadable text, or identity drift.
A good Seedance prompt should read like a director’s note, not like a keyword list.
Weak prompt:
Make a cool video of a man walking back home tiredly, cinematic.Better prompt:
提示：图片 1 中的男子下班后疲惫地走在走廊上，放慢脚步，最后停在大门口。面部特写，他深吸一口气，调整情绪，收起负面情绪，变得轻松。然后是一个特写镜头，他翻找钥匙，将钥匙插入门锁，进入家门。他的小女儿和一只宠物狗高兴地跑过来迎接和拥抱他。室内非常温馨舒适，对话自然流畅。.
How to Write Seedance Prompts for Different Input Types
Different Seedance workflows require different prompt styles.
| 提示类型 | 您的输入 | 最佳使用案例 | 何时使用 |
| T2V (Text-to-Video) | 仅限文本 | Complex physical movements, sports, fixing gravity glitches. | When you need realistic action from scratch without any image assets. |
| I2V (Image-to-Video) | 1 张图片 + 文字 | Micro-expressions, product commercials, cinematic ASMR videos (focusing on texture & touch). | When you want a specific picture to come alive and interact with objects. |
| R2V (Reference-to-Video) | 多个文件（@ 标记） + 文本 | Multiverse transitions, storyboards, exact style/character copying. | When you need strict consistency or complex multimodal logic across scenes. |
| A2V (Audio-to-Video) | Audio + Image/Video + Text | Lip-syncing, rhythmic transitions, high-impact sound design, or music videos. | When the movement must be perfectly synced to the rhythm or timing of a sound file. |
Text-to-Video Prompt
Use text-to-video when you do not need a specific character, product, or reference motion.
最适合
- abstract concepts
- mood videos
- background scenes
- generic cinematic shots
- first-round ideation
模板：
Create a video of [subject] [action] in [scene].
Camera: [shot type and movement].
Style: [lighting, color, mood, genre].
Details: [materials, atmosphere, background elements].
Quality rules: sharp focus, realistic motion, no distortion, no unreadable text.示例 1：
Competitive pair figure skating scene. Opening with a low-angle tracking shot following the ice blades, clear ice shavings and reflection details. Entering the spin segment, the male skater's axis slightly shifts causing a mistake, the spin rhythm briefly collapses. The female skater quickly adjusts her center of gravity, with a calm expression signaling 'Stay with me', actively guiding the male skater to realign the rhythm. Then seamlessly transitions into a lift, clean and stable lines. The climax is a synchronized jump combination, straight aerial posture, decisive landing, audio and visual perfectly aligned. The female skater wears a dark blue figure skating dress, the male skater is in competitive sportswear示例 2：
您无需随后添加声音。该人工智能 创建本地双声道音频 与视频完美匹配。它甚至还能发出高质量的 ASMR 声音。.
Immersive first-person perspective hand ASMR video. Close-up shot, under warm soft light, a pair of slender hands gently triggers different objects in sequence: the light scraping of frosted glass, the rubbing of plush fabric, the light tapping of an acrylic board, the light squeezing of bubble wrap, the light scratching of a wooden comb. Fingers move slowly and gently, no background music, pure natural trigger sounds, relaxed and healing visual atmosphere."Example3:
武侠风格的视听大片，白衣剑客与竹林侠客的对峙。镜头在两人之间缓缓移动，焦点在雨滴和剑柄之间切换。现场气氛异常压抑，只能听到 "哗哗 "的雨声。突然，一道雷光闪过，两人同时冲了上去。侧拍镜头高速移动，捕捉到泥水飞溅的脚步声。两兵相遇的瞬间，画面切换到极慢动作，清晰地呈现出剑气震雨形成的环形冲击波，以及被剑气斩断的竹叶。随后，恢复正常速度，两人背靠背落地，剑客斗笠破裂，画面戛然而止。.Image-to-Video Prompt
Use image-to-video when identity or product accuracy matters.
最适合
- character animation
- product ads
- logo reveals
- fashion motion
- two-photo short films
模板：
Use @Image1 as the exact visual reference for [subject/product/logo].
Animate it by [action].
The scene is [environment].
Camera: [movement].
Style: [visual direction].
Keep [key visual details] consistent throughout.示例 1：
Use @Image1 as the exact visual reference for the main character. She feels guilty, looks left and right, reaches out of the frame, grabs a cola and takes a sip, showing a satisfied expression. Hearing footsteps, the character quickly puts the cola back. At this moment, a cowboy walks in and takes the cola. The ending zooms into a top-lit close-up of the cola against a pure black background, with artistic subtitles and voiceover: 'Yikou Cola, you must taste it!'Example2：
20 世纪 20 年代爵士俱乐部风格的查尔斯顿舞。一位身着金色流苏裙的女舞者和一位身着条纹西装的男舞者表演高强度动作。动作包括快速的切分音步、空中抛接和大臂横扫。镜头采用动态跟踪拍摄，穿插脚部动作特写。重点是身体细节，如每一脚踢出时狂乱甩动的流苏、皮肤上闪烁的汗珠，以及带有烟雾效果的复古胶片颗粒纹理。背景爵士乐队和欢呼的观众营造出一种强烈的派对气氛。.Example 3:
A girl in a hat in the middle sings softly, "I'm so proud of my family!" and then turns to embrace the black girl in the center. The black girl responds emotionally, "My sweetie, you're the heart of our family," and returns the embrace. The boy in yellow clothes on the left says happily, "Folks, let's dance together to celebrate!" The girl on the far right replies immediately, "I'll bring the music!" as the background Latin music starts playing. The woman in an orange dress on the left (Julietta) smiles and nods, while the woman on the right with braids (Luiza) clenches her fists and waves her arms. Some people in the crowd begin to step their feet, the children clap in rhythm, and the whole family is about to form a circle, dancing joyfully to the upbeat music, their skirts fluttering, expressing happiness and warmth on the colorful streets.
Video-to-Video Prompt
Use video-to-video when movement matters more than description.
最适合
- viral video replication
- dance motion transfer
- camera motion transfer
- action reference
- fashion walking sequences
模板：
Use @Video1 as the motion and camera reference.
Use @Image1 as the new subject reference.
Keep the camera movement, pacing, and main action from @Video1, but replace the original subject with @Image1.
Generate a new scene in [environment/style].
Do not copy the original background, face, or branding from @Video1.Audio-to-Video Prompt
Use audio-to-video when timing, rhythm, or sound-on performance matters.
最适合
- music videos
- dancing
- ad pacing
- 对话
- voiceover
- emotional scenes
模板：
Use @Audio1 as the rhythm and mood reference.
Create a video where [subject] [action] in sync with the beat / voice / sound.
Camera cuts and movement should follow the pacing of @Audio1.
Style: [visual style].
Keep the subject consistent and make all motion natural.例如
Use @Audio1 as the rhythm reference.
Create a fast 9:16 sneaker ad where the sneaker from @Image1 lands on a reflective floor exactly on the first beat.
Each beat triggers a new shot: sole close-up, side profile, lace detail, final hero pose.
High-energy sports commercial, dark background, sharp rim light, clean reflections.
Keep the sneaker shape, color, logo, and material unchanged.
5 Copy-Paste Prompts
提示 1：动漫女孩治愈入口
Seedance 2.0 提示：
在夏日午后温暖的阳光下，一位 18 岁的日本短发动漫女孩身穿白色连衣裙，头戴草帽，站在林间小道上。她慢慢地转向镜头，轻轻地微笑着。一阵微风吹动着她的头发和裙子。镜头从中景慢慢推进到特写。柔和的自然光、胶片颗粒、治愈平和的气氛、电影质感。保持面部和服装的一致性，无失真，细节丰富。.
参考资料：上传一张清晰的正面动漫人物图像。.
提示 2：产品 360° 旋转（电子商务）
Seedance 2.0 提示：
在纯白的无限工作室背景上，一个极简的黑色哑光机械键盘顺时针平滑旋转 360 度。RGB 灯光轻柔呼吸。键帽文字清晰可读。固定微距相机，平滑的转盘运动，商业产品摄影风格，柔和的高调照明，无噪音。徽标和文字保持完全一致。.
参考：上传高分辨率产品图片。.
提示 3：动作打斗场景（视频参考）
Seedance 2.0 提示：
武侠风格的男主角（根据参考视频角色改编）身着黑色武侠装束，在雨夜的竹林中与敌人搏斗。快剑连击，剑光轨迹清晰可见，水花四溅。快速跟拍、吊臂镜头和快速特写。电影镜头语言。保持角色外观和服装的一致性。逼真的物理效果、湿布料和雨水互动。.
参考资料：上传一段武术视频和人物形象。.
提示 4：音乐视频节拍同步（音频驱动）
Seedance 2.0 提示：
一个时髦的赛博朋克女孩在夜晚霓虹闪烁的城市街道上翩翩起舞。每一个强烈的节拍都会引发剪切或快速的镜头移动。霓虹灯在潮湿的地面上反光。赛博朋克风格，快节奏剪辑，多镜头连续性。舞蹈动作和人物外观保持一致。.
参考资料：上传 BGM 音频 + 舞蹈参考视频或图片。.
提示 5：多镜头讲故事
Seedance 2.0 提示：
一个孤独的机器人在一座废弃的工厂里醒来（场景 1）。它走到外面，看到一片夕阳下的荒地（场景 2）。它发现了一朵小花，轻轻地抚摸着它（场景 3）。最后，它抬起头，对着天空微笑（场景 4）。保持机器人外观的一致性。从困惑到温暖的情感过渡。电影镜头、暖色调、史诗氛围、无闪烁。.
参考资料：上传机器人设计 + 日落荒原图片。.
Mastering the Art of the Prompt: 7 Rules for Seedance 2.0 Success
Even with the most powerful AI, the quality of your output is determined by the clarity of your input. To help you bypass “AI hallucinations” and common artifacts, we have compiled the ultimate Seedance 2.0 Prompting Blueprint.
Follow these seven professional standards to turn your vision into a pixel-perfect commercial.
1. Assign “Roles” to Your @Mentions
Never mention your files casually. Start your prompt by defining exactly what each input controls. This anchors the AI’s logic.
- 例如 “Use @Image1 for identity, @Video1 for motion, and @Audio1 for rhythm.”
2. The “One Action” Rule
AI video models struggle with complex choreography. To prevent “melting” or chaotic motion, limit each clip to one primary action.
- 很好： “The character turns around and smiles.”
- 糟糕： “The character runs, jumps, opens a door, and then flies away.”
3. Speak the Language of Cinema
Seedance 2.0 responds best to professional camera direction. Instead of saying “zoom,” use specific filmmaker terminology:
| Camera Term | Meaning | 如何提示 |
| Wide shot | Shows full scene and environment | “A cinematic wide shot of the entire landscape.” |
| Medium shot | Subject from waist/chest up | “A medium shot showing the character from the waist up.” |
| Close-up | Face or object detail | “A sharp close-up focusing on the product texture.” |
| Dolly-in | Camera moves closer to subject | “A smooth dolly-in toward the product on the table.” |
| Tracking shot | Camera follows subject | “A high-speed tracking shot following the car from behind.” |
| Pan left/right | Camera rotates horizontally | “Slowly pan right to reveal the interior of the room.” |
| Tilt up/down | Camera rotates vertically | “Tilt up from the character’s feet to their face.” |
| Handheld | Natural human camera shake | “A handheld shot with subtle organic movement.” |
| Static shot | Fixed camera, no movement | “A static shot with zero camera movement, locked focus.” |
4. The Power of Repetition
For commercial consistency, redundancy is your friend. If a detail must stay 100% the same, repeat it in your “Lock Block”:
“Keep the same face, same hairstyle, same outfit, and same body proportions throughout. No flickering.”
5. Design for “Post-Production”
A common mistake is trying to generate perfect text within the AI video. AI models often distort typography.
- The Pro-Way: Generate a clean visual “Hero Shot” first, then add your slogans, subtitles, and UI copy using professional editing software.
6. “Sprint” Before You “Marathon”
Don’t waste credits on long, 15-second generations until you are sure the prompt works.
- Test Phase: Generate 4–5 second “Short Tests” to check identity and motion.
- Production Phase: Once the prompt is stable, generate the full-length version.
7. Use “Shot-by-Shot” Prompting
For complex narratives, treat your prompt like a storyboard. Break the video into distinct shots:
- 镜头 1： Macro close-up of texture.
- 镜头 2： Wide-angle reveal.
- 镜头 3： High-energy hero shot.
Advanced Troubleshooting Cheat Sheet
| To Fix This… | Do This… |
| Motion Artifacts | Change “fast” to “medium speed” or use a @Video motion reference. |
| Hand/Face Distortion | 添加： “Hands with perfect anatomy, clear facial features, no flickering.” |
| Randomness/Drifting | Upload more reference images (up to 9) to “lock” the visual environment. |
| Long-Form Narrative | Generate a perfect 5s clip, then use the 扩展视频 feature to continue. |
为什么 Seedance 2.0 的提示感觉就像在导演一部电影
Seedance 2.0 将人工智能视频创作提升到了一个新的高度。现在，创作者可以近乎导演般精确地指导角色、摄像机、动作和声音，而不是与随机输出进行对抗。.
有了正确的 Seedance 2.0 提示结构, 您可以从一个简单的想法变成一个电影般的、高留存率的人工智能视频，而且看起来是专业制作的。.
2026 年最新更新：在哪里不受限制地执行您的提示
As of 2026, the biggest challenge for AI directors isn’t writing the perfect prompt—it’s getting past the login screen. While Seedance 2.0 remains heavily gated behind Chinese regional verifications and strict platform constraints, and other official platforms demand steep $200/month paywalls or face sudden unavailability, your highly crafted prompt templates are useless without a stable rendering engine.
您不再需要浪费时间寻找临时测试版账户、绕过 IP 屏蔽或等待邀请码。对于创作者来说，目前最有效的策略就是在统一平台上部署这些精确的导演式提示，以保证不间断的跨模式访问。.
