Seedance 2.0 — Cinematic-Grade AI Video Generator by ByteDance
The complete prompting guide for Seedance 2.0. Learn image-to-video and text-to-video techniques to generate cinematic AI videos with multi-shot control, native audio, and 2K resolution.
Seedance 2.0 is ByteDance's cinematic-grade AI video generator. Create studio-quality videos from images or text with multi-shot generation, native audio sync, 2K resolution, and director-level camera control — running 6x faster on WaveSpeed's optimized infrastructure.
Prompting Guide — Table of Contents
Image-to-Video Techniques
Text-to-Video Techniques
Image-to-Video TechniquesBasic Prompt Structure
Prompt = [subject] + [Motion] , [Background] + [Motion], [Camera] + [Motion]...
1. Keep it simple and direct
Use simple, clear language. Seedance 2.0 intelligently expands your prompt based on the input image, so concise descriptions produce the best results.
2. Negative prompts are not supported
Seedance 2.0 does not process negative prompts. Instead of saying what you don't want, describe exactly what you do want.
3. Focus on motion, not static elements
For image-to-video, the scene already exists in your input image. Focus your prompt on what should move — the subject's action, background changes, and camera movement — rather than describing static elements that are already visible.
4. Highlight distinctive features
When your subject has distinctive features, mention them to help the model identify the correct subject. For example: 'an elderly man with a white beard' or 'a woman wearing red sunglasses'. For motion, always specify intensity with clear adverbs like 'quickly' or 'gently'.
5. Stay consistent with the input image
Your prompt must be consistent with the input image. If the image shows a man, don't prompt for a woman dancing. If the background is a meadow, don't describe a cafe scene. If no jewelry is visible, don't reference jewelry. If you select a fixed camera, don't write camera orbit instructions.
Multi-Action Sequences
Seedance 2.0 excels at multi-action sequences. It supports multiple continuous actions in temporal order and coordinated actions across different subjects within a single generation.
Prompt word = Subject 1 + Movement 1 + Movement 2
Prompt word = Subject 1 + Movement 1 + Subject 2 + Movement 2 ...
List each action sequentially. The model interprets the temporal flow and generates smooth transitions between movements automatically.
Camera Movement Control
Seedance 2.0 offers director-level camera control via natural language. Describe the camera behavior you want — orbit, aerial, zoom in/out, pan, tracking shot, handheld shake — and the model will execute it. Multi-shot transitions with natural cuts are also supported.
Intensity & Degree Modifiers
Adverbs of degree control the intensity, speed, and amplitude of actions in your generated video. Without explicit intensity cues, the model defaults to its own interpretation — which may not match your intent.
Key Principles
1. Be explicit about intensity. The model cannot infer motion intensity from a static reference image. Instead of 'the car drove by', write 'the car drove by at high speed'. Specificity produces more accurate results.
2. Exaggerate for impact. Changing 'man roaring' to 'man roaring furiously' or 'wings flapping' to 'wings flapping vigorously' dramatically improves the expressiveness of the generated video.
Intensity modifier examples:
Text-to-Video TechniquesBasic Prompt Structure
Prompt = [subject] + [Movement] + [Scene] + [Lens], [Style]...
* Subject + Motion + Scene are the core elements of every text-to-video prompt. The model expands your description and generates a video that matches your intent.
* All image-to-video guidelines — multi-action sequences, camera movement, intensity modifiers — apply equally to text-to-video generation. Negative prompts are not supported in either mode.
How to Write Better Prompts:
1. Detailed character description
Specify the character's appearance, clothing, hairstyle, posture, and expression. The more visual detail you provide, the more accurate and consistent the generated character will be.
2. Environment and setting
Describe the environment in sensory detail — mountain peaks, desert dunes, waterfalls, neon-lit streets, or dimly lit studios. Rich environmental descriptions ground the video in a believable visual context.
3. Emotion and dynamic interaction
Combine character emotions with environmental dynamics to create narrative depth. A character's expression, body language, and interaction with the surroundings bring the scene to life.
4. Atmosphere and lighting
Lighting sets the mood. Use descriptive terms like 'golden hour sunlight', 'overcast dawn', 'warm candlelight', or 'harsh neon glow' to control the atmosphere of your generated video.
Using Cases
Cinematic Storytelling: A cinematic film production scene with a director's chair, clapperboard, and multiple camera angles showing a dramatic story unfolding.
Music Video: A vibrant music video production scene with a performer on a neon-lit stage with dynamic light beams and synchronized visual effects.
Product Showcase: A premium product showcase with a luxury item rotating in cinematic slow motion with professional studio lighting.
Nature Documentary: An epic nature documentary scene with a majestic eagle soaring over snow-capped mountains with dramatic clouds.
Social Media Content: A modern social media content creation setup with a ring light, smartphone on tripod, and colorful backgrounds.
Animation & Effects: A magical animation and visual effects scene with abstract particles transforming into characters and creatures.