Seedance 2.0 Prompt Guide: From Basics to Mastery

Seedance 2.0 natively supports joint audio-video generation with exceptional semantic understanding and multimodal interaction capabilities. This guide covers prompt techniques for Seedance 2.0 (also applicable to Seedance 2.0 Fast) to help you create high-quality videos efficiently.

1. Core Principles

The Prompt Formula

Seedance 2.0 deeply follows natural language logic. Build your prompts by combining these elements:

Required Elements:

Optional Elements:

In short: Start with "who is doing what," then add "where, what style, how to shoot, what sounds."

Multimodal Reference Control

Beyond text, you can upload reference materials (images, videos, audio) for precise control:

• Be explicit: Clearly specify reference targets in your prompt, e.g., "reference the composition of Image 1" or "follow the motion in Video 2"
• Precise transfer: The model automatically extracts core features from references and combines them with your text description

2. Text Generation in Videos

Seedance 2.0 can generate on-screen text in text-to-video, image-to-video, and reference-to-video workflows. The model automatically matches appropriate text styles and colors, and also supports manual specification.

Use common characters for best results — avoid rare characters and special symbols.

Slogans

Template:

"Text content" + "when it appears" + "where it appears" + "how it appears", "text style (color, font)"

Example:

Hand-drawn comic style, three people sitting together eating fried chicken from Image 1, friendly and joyful atmosphere, then the scene gradually blurs, text "Joy is in Seedance" appears in the center of the screen.

Subtitles

Template:

Subtitles appear at the bottom of the screen, reading "...", synced with the audio rhythm.

Example:

Generate a video with voiceover. A deep, calm male voice says: "In the vast universe, our world is but a brief moment. Yet within it, life flourishes against all odds." The scene transitions slowly from night to dawn, stars fading, sun rising behind mountains. Subtitles appear at the bottom following the narration.

Speech Bubbles

Template:

"Character" says: "...", speech bubbles appear around the character with the dialogue text.

Example:

Two people from Image 1 wearing sportswear running on a school track. The girl looks at the boy and confidently says: "We can definitely do it!" Cut to a close-up of the boy, who hesitantly replies: "Are you sure?" Cut back to the girl in mid-close-up, she says cheerfully: "Yes!" Bright and determined mood. Speech bubbles appear around each speaking character with corresponding dialogue.

3. Image References

Seedance 2.0 supports multi-angle subject references and multi-image references. Upload images in order, and use Image 1, Image 2... Image n in your prompts.

Multi-Angle Subject Reference

Best for maintaining subject consistency — product showcases, character animations, etc.

Template:

Reference/Extract/Combine + "Image n" + "subject", generate "scene description", keep "subject" features consistent.

Product showcase example:

Extract the camera from Image 1, Image 2, Image 3, change background to white, camera on a white table, lens focuses on the camera in close-up, then slowly rotates around the camera as the subject, clearly showing front, side, and back.

Character reference example:

Reference the woman in Image 1, Image 2, Image 3, generate a scene of her eating cake in a coffee shop.

Tip: Multi-angle reference photos (front, side, back) help the model reproduce subject features more accurately.

Multi-Image Reference

For complex scenes combining multiple different elements.

Template:

Reference/Extract/Combine/Follow + "Image n" + "referenced element", generate "scene description", keep "referenced element" features consistent.

Logo reference example:

Background is a neon-lit futuristic city aerial corridor with flying vehicles and holographic ads. Reference the girl in Image 2, first show her in mid-shot releasing silver floating lanterns with holographic projections, then pull back to show floating lanterns filling the sky, scene gradually blurs, then Image 1's Logo appears. Overall style: 3D cyberpunk sci-fi animation.

Multi-element reference example (character + clothing + scene + logo):

Scene set in the restaurant from Image 4, busy with customers. The girl from Image 1 wearing the outfit from Image 2 is organizing items on the counter. The boy from Image 3 is a customer who walks up to ask for her contact info. The logo from Image 5 is always displayed in the bottom-right corner.

Storyboard reference example:

Follow the storyboard in the image, generate an intense fight scene. Each panel's composition should appear in sequence, then the two engage in fierce combat.

Tip: Each image plays a different role — character, clothing, scene, logo, etc. Clearly state each image's purpose in the prompt.

4. Video References

Seedance 2.0 can extract motion, camera work, and effects from reference videos. Upload in order and use Video 1, Video 2... Video n.

Motion Reference

Extract motion details from a reference video and apply to new characters or scenes.

Template:

Reference "Video n"'s "motion description", generate "scene description", keep motion details consistent.

Film example:

Reference Video 1's character movements and cinematography, generate a fight scene between Image 2 and Image 1, Image 2 is the left character, Image 1 is the right character. Intense background music.

Marketing example:

Reference the horse's running form in Video 1, generate a golden horse galloping across grasslands, then freeze its magnificent running pose, transforming into a horse-shaped gold pendant.

Camera Movement Reference

Extract camera movement from a reference video and apply to entirely new scenes.

Template:

Reference "Video n"'s "camera movement", generate "scene description", keep camera movement consistent.

Example:

Reference Video 1's camera movement, create a tech park concept video with the skyscraper from Image 1 as the visual center, same first-person diving perspective, showcasing the tech aesthetic of Image 1's campus.

Effects Reference

Extract visual effects from reference videos and apply to new scenes.

Template:

Reference "Video n"'s "effect description", generate "scene description", keep effects consistent.

Film effects example:

Reference Video 1's golden particle effects, have the character from Image 2 playing flute while surrounded by the same particle effects.

Creative effects example:

Reference Video 1's effects, have the girl from Image 1 grow the same wings, with matching wing generation trajectory.

5. Video Editing

Seedance 2.0 supports editing existing videos — adding, removing, or modifying elements, extending videos, and filling gaps between clips.

Add / Remove / Modify Elements

Templates:

Add: Add "desired element" at "time/spatial position" in "Video n".
Remove: Remove "element" from "Video n", keep other content unchanged.
Modify: Replace "element" in "Video n" with "new element".

Add example:

Add fried chicken, pizza and other snacks on the table in Video 1.

Remove example:

Clear all extra parts and tools from the desk in Video 1, keep the desk clean with only what they're holding.

Modify example:

Replace the perfume in Video 1 with the face cream from Image 1, keep motion and camera movement unchanged.

Tip: When modifying elements, adding "keep motion and camera movement unchanged" maintains video continuity.

Video Extension

Extend existing videos forward or backward. The model automatically handles transitions.

Templates:

Extend "Video n" forward/backward + "description of extended content"
Generate content before/after "Video n" + "description"

Extend backward example:

Generate content after Video 1, two late-arriving men running toward them, all five finally meeting and chatting happily.

Extend forward example:

Extend Video 1 forward, over-the-shoulder shot of the man in white, he says: "It's not that bad. You're just stressed. Everyone goes through this, you just need to keep going."

Note: Original footage is not re-generated — the model seamlessly connects new content to existing clips.

Track Filling (Gap Bridging)

Seamlessly connect multiple video clips with transition scenes.

Template:

"Video 1" + "transition description" + then "Video 2" + "transition description" + then "Video 3"

Example:

Video 1, the moment the leaf touches the ground, golden particle effects burst out, a gust of wind blows, then Video 2.

Limitations:

• Maximum 3 video inputs, total duration under 15 seconds
• The model automatically trims connecting portions from each clip

6. Best Practices

5 Principles for Better Prompts

1. Clear subject: Always start with "who is doing what" — this is the foundation
2. Rich details: Add environment, lighting, style, and mood — more details = more control
3. Explicit references: Use Image 1, Video 2, etc. to clearly identify uploaded materials
4. Consistent ordering: Number materials by upload order to avoid confusion
5. Natural language: Write naturally — avoid keyword stuffing

Common Prompt Structures

Basic text-to-video:

"Subject + action" + "environment/background" + "visual style" + "camera movement" + "audio"

Image-referenced video:

Reference "Image n"'s "element", generate "scene description", keep "element" features consistent.

Video-referenced video:

Reference "Video n"'s "motion/camera/effects", generate "new scene", keep "referenced content" consistent.

Video editing:

Add/Remove/Replace "target element" in "Video n", "edit description", keep other content unchanged.

Start with simple text-to-video prompts, then gradually explore image references, video references, and video editing to unlock the full potential of Seedance 2.0.