Generate and direct cinematic AI videos with Seedance 2.0 (ByteDance / Dreamina / Jimeng).
Text-to-Video Β· Image-to-Video Β· Video-to-Video Β· Reference-to-Video
Author: Emily (@iamemily2050) Β |Β
π Β |Β
IG
Platform: ByteDance Seedance 2.0 Β· Dreamina Β· Jimeng
Updated: 2026-04-27 Β· v5.1.0 validation and source-hygiene release
This library teaches you to direct the AI, not micro-manage it. Tell the model WHAT you want and HOW it should FEEL. Use @references to show, not tell.
| Workflow | Best for | Start here |
|---|---|---|
| Full Interview | Vague idea, need creative guidance | seedance-interview |
| Direct Prompt | Clear vision, have reference media | seedance-prompt |
Note from the field (verified 2026-04-27): Seedance V2 performs significantly better with short prompts (30-100 words) written in Chinese. The v5.1 prompt system enforces this by default.
|
For AI Filmmakers β Stop writing flat prompts. Seedance 2.0 gives you a complete director's toolkit: camera language, motion control, lighting design, character fidelity, audio sync, and VFX integration β all structured as modular, composable skills that any AI agent can load on demand. |
For Agent Builders β Each of the 23 sub-skills is independently loadable. Your agent reads the root |
β οΈ Status (verified 2026-04-27): Seedance 2.0 platform and API behavior changes quickly. For API availability, face/portrait authorization, upload limits, pricing, and regional access, checkreferences/api-status.mdand verify current primary sources. Real-person likeness workflows require authorization and are surface-specific. Runseedance-copyrightbefore protected-IP, celebrity, brand, or real-person workflows.
Click any node to navigate directly to that skill's documentation.
π Full Skill Directory β Core Pipeline
| Skill | Emoji | What it does |
|---|---|---|
seedance-interview |
π | Director's Journey with Quick Mode exit and genre detection |
seedance-interview-short |
ποΈ | Compressed interview outputting a 30-100 word brief with live counter |
seedance-prompt |
βοΈ | Director's Formula: genre router, 30-100 word target, I2V gate, anti-slop check |
seedance-prompt-short |
β‘οΈ | 30-100 word budget system with Compression Engine |
seedance-camera |
π₯ | One-Move Rule, genre camera presets, camera transfer via @Video |
seedance-motion |
π | Intent-first motion, @Video reference primary, physics consequences |
seedance-lighting |
π‘ | Lighting, atmosphere, light transitions, mood and time-of-day specs |
seedance-characters |
π | Character identity locking, @Tag assignment, multi-character scene management |
seedance-style |
π¨ | Visual style, render-engine tokens, period aesthetics, style-transfer reference |
seedance-vfx |
β¨ | VFX physics contracts, particle systems, destruction, energy effects |
seedance-audio |
π | Native audio design, dialogue lip-sync, @Audio1 reference, desync fixes |
seedance-pipeline |
π | ComfyUI nodes, API integration, Firebase Studio, post-processing chains |
seedance-recipes |
π | 7-genre template library: product, lifestyle, drama, MV, landscape, commercial, anime |
seedance-troubleshoot |
π§ | Diagnostic tree: 5-step root cause analysis for common failure modes |
βοΈ Content Quality & Governance
| Skill | Emoji | What it does |
|---|---|---|
seedance-copyright |
βοΈ | IP rules, safe substitutions, authorization-dependent enforcement guidance |
seedance-antislop |
π« | Detects and removes AI filler language and hollow superlatives from prompts |
seedance-filter |
π‘οΈ | Content filter intelligence: diagnose false-positive blocks, write prompts that pass |
π Multilingual Vocabulary
| Skill | Flag | Languages | Terms |
|---|---|---|---|
seedance-vocab-zh |
π¨π³ | Chinese cinematic vocabulary | 400+ |
seedance-vocab-ja |
π―π΅ | Japanese cinematic vocabulary | 280+ |
seedance-vocab-ko |
π°π· | Korean cinematic vocabulary | 270+ |
seedance-vocab-es |
πͺπΈ | Spanish cinematic vocabulary (Castilian + Latin American) | 270+ |
seedance-vocab-ru |
π·πΊ | Russian cinematic vocabulary (Eisenstein/Tarkovsky tradition) | 270+ |
π¨π³ Working Examples (Chinese Prompts)
| # | Genre | Difficulty | Description |
|---|---|---|---|
| 1 | ε§ζ ηε§ Short Drama | Expert | ιΈιζ»θ£η½ε§ι£ζ Ό β 15s multi-shot reversal scene |
| 2 | ε§ζ ηε§ Short Drama | Beginner | δΌι ζΎθ‘£εΊζ― β simple elegant action loop |
| 3 | ε§ζ ηε§ Short Drama | Intermediate | η»΄ε€ε©δΊζΆδ»£θ‘ζ― β period drama environment |
| 4 | εΉΏε Advertising | Advanced | δΊε¨η»η»θ§θ² β painting character comes alive |
| 5 | εΉΏε Advertising | Creative | ζ©ζ车广ε β donkey motorcycle stunt ad |
| 6 | εΉΏε Advertising | Creative | ε转ιΆι£εΉΏε β spy thriller snack reveal |
| 7 | ε¨ζΌ«ζ¦ζ Animation | Expert | εͺε vs ζδΈ β 4-act ice/fire battle sequence |
| 8 | ε¨ζΌ«ζ¦ζ Animation | Advanced | ε€θ§ι’εθζζ β multi-reference fight scene |
| 9 | δΊ§εε±η€Ί Product | Intermediate | ι«η«―ι¦ζ°΄ MG ε¨η» β luxury perfume ad |
| 10 | δΊ§εε±η€Ί Product | Advanced | ε€εΎθεδΊ§εε±η€Ί β multi-image bag showcase |
| 11 | δΊ§εε±η€Ί Product | Advanced | η»ε ΈεΉΏεθε₯ε€ε» β car ad licensed rhythm reference |
| 12 | θ§θ§ηΉζ VFX | Advanced | η²εηΉζη倴 β gold particle title animation |
| 13 | θΏιεδΊ Cinematography | Advanced | δΈιε°εΊθΏ½θΈͺι倴 β one-take tracking shot |
| 14 | θΏιεδΊ Cinematography | Advanced | ε¨δ½+θΏιειε€ε» β dance licensed performance reference |
| 15 | θΏιεδΊ Cinematography | Advanced | θ§θ²ζΏζ’ β character replacement in video |
| 16 | ι³δΉε‘ηΉ Beat Sync | Advanced | ι£ε ηι³δΉε‘ηΉ β landscape beat sync |
β Full prompts: skills/seedance-examples-zh/SKILL.md
# Antigravity
antigravity skills install https://github.com/Emily2040/seedance-2.0
# Gemini CLI
gemini skills install https://github.com/Emily2040/seedance-2.0
# Claude Code
claude skills install https://github.com/Emily2040/seedance-2.0
# GitHub Copilot / Codex
codex skills install https://github.com/Emily2040/seedance-2.0
# Cursor
cursor skills install https://github.com/Emily2040/seedance-2.0
# Windsurf
windsurf skills install https://github.com/Emily2040/seedance-2.0
# OpenCode
opencode skills install https://github.com/Emily2040/seedance-2.0π Manual Installation Paths
| Platform | Workspace path | Global path |
|---|---|---|
| Antigravity | .agent/skills/seedance-20/ |
~/.gemini/antigravity/skills/seedance-20/ |
| Gemini CLI | .gemini/skills/seedance-20/ |
~/.gemini/skills/seedance-20/ |
| Firebase Studio | .idx/skills/seedance-20/ |
β |
| Claude Code | .claude/skills/seedance-20/ |
~/.claude/skills/seedance-20/ |
| OpenClaw | .claude/skills/seedance-20/ |
~/.claude/skills/seedance-20/ |
| GitHub Copilot | .github/skills/seedance-20/ |
~/.copilot/skills/seedance-20/ |
| Codex | .agents/skills/seedance-20/ |
~/.agents/skills/seedance-20/ |
| Cursor | .cursor/skills/seedance-20/ |
~/.cursor/skills/seedance-20/ |
| Windsurf | .windsurf/skills/seedance-20/ |
~/.codeium/windsurf/skills/seedance-20/ |
| OpenCode | .opencode/skills/seedance-20/ |
~/.config/opencode/skills/seedance-20/ |
| Antigravity | Gemini CLI | Firebase Studio | Claude Code | OpenClaw | Copilot | Codex | Cursor | Windsurf |
| β | β | β | β | β | β | β | β | β |
Seedance 2.0 includes dedicated cinematic vocabulary modules for five languages, enabling native-language prompt engineering for maximum generation fidelity.
π¨π³ Chinesevocab-zh400+ terms |
π―π΅ Japanesevocab-ja235 lines |
π°π· Koreanvocab-ko225 lines |
πͺπΈ Spanishvocab-es232 lines |
π·πΊ Russianvocab-ru235 lines |
seedance-2.0/
βββ SKILL.md β Root entry point (61 lines)
βββ LICENSE β MIT
βββ README.md β You are here
βββ CHANGELOG.md β v3.0.0 β v5.1.0
βββ .github/
β βββ CODEOWNERS β @Emily2040
βββ skills/ β 23 modular sub-skills
β βββ seedance-interview/ β π Director's Journey + Quick Mode
β βββ seedance-interview-short/ β ποΈ Compressed interview (30-100 words)
β βββ seedance-prompt/ β βοΈ Director's Formula + Genre Router
β βββ seedance-prompt-short/ β β‘οΈ Compressed prompt (30-100 words)
β βββ seedance-camera/ β π₯ Camera language
β βββ seedance-motion/ β π Motion control
β βββ seedance-lighting/ β π‘ Lighting design
β βββ seedance-characters/ β π Character fidelity
β βββ seedance-style/ β π¨ Style control
β βββ seedance-vfx/ β β¨ VFX integration
β βββ seedance-audio/ β π Audio & lip-sync
β βββ seedance-pipeline/ β π Pipeline ops
β βββ seedance-recipes/ β π Genre recipes
β βββ seedance-troubleshoot/ β π§ Diagnostic Tree
β βββ seedance-copyright/ β βοΈ IP governance
β βββ seedance-antislop/ β π« Language filter
β βββ seedance-vocab-zh/ β π¨π³ Chinese
β βββ seedance-vocab-ja/ β π―π΅ Japanese
β βββ seedance-vocab-ko/ β π°π· Korean
β βββ seedance-vocab-es/ β πͺπΈ Spanish
β βββ seedance-vocab-ru/ β π·πΊ Russian
β βββ seedance-filter/ β π‘οΈ Content filter intelligence
β βββ seedance-examples-zh/ β π¨π³ Chinese Working Examples
βββ references/ β 14 reference files
βββ platform-constraints.md β Platform limits & rules
βββ json-schema.md β JSON prompt schema
βββ prompt-examples.md β Copy-paste examples
βββ storytelling-framework.md β Narrative design & visual layering principles
βββ quick-ref.md β Quick reference card
βββ genre-guides.md β 7-genre prompt templates
βββ reference-workflow.md β @reference system guide
βββ i2v-guide.md β Image-to-Video best practices
βββ intent-vs-precision.md β Intent-first prompting philosophy
βββ api-status.md β Current API and platform status
βββ source-registry.md β Preferred sources for factual claims
βββ audio-guide.md β Audio layer and lip-sync patterns
βββ anti-slop-lexicon.md β Weak phrase replacement table
βββ filter-vocab.md β Filter-safe vocabulary substitutions
| File | Purpose |
|---|---|
platform-constraints.md |
Platform limits, resolution caps, and known behaviors |
json-schema.md |
JSON prompt schema for structured generation |
prompt-examples.md |
Compact, copy-paste prompt examples |
quick-ref.md |
Quick reference card for all parameters |
storytelling-framework.md |
Narrative design, visual layering, and director's toolkit principles |
genre-guides.md |
7-genre prompt templates and best practices |
reference-workflow.md |
The @reference system: show, don't tell |
i2v-guide.md |
Image-to-Video best practices |
intent-vs-precision.md |
Intent-first prompting philosophy |
api-status.md |
Current API and platform status (dated, source-aware) |
source-registry.md |
Preferred sources for factual claims |
audio-guide.md |
Audio layer and lip-sync patterns |
anti-slop-lexicon.md |
Weak phrase replacement table |
filter-vocab.md |
Filter-safe vocabulary substitutions |
All 23 skills pass the AgentSkills open standard validation:
- β
nameβ lowercase, hyphen-separated, no dots or spaces - β
descriptionβ single-quoted, verb-first, includes WHEN trigger phrases - β
license: MITΒ·user-invocable: trueΒ·user-invokable: true - β
tags:array andmetadatawithversion,updated,author, platform blocks - β
metadata.parent: seedance-20on all 23 sub-skills - β No illegal top-level custom fields
See CHANGELOG.md for the full version history from v3.0.0 to v5.1.0.
Contributions are welcome. Fork the repository, create a feature branch, and submit a pull request. All contributions will be reviewed by @Emily2040.
MIT Β© 2026 Emily (@iamemily2050)
Built with precision by Emily (@iamemily2050) β AI artist, filmmaker, and agent skill architect.
π @iamemily2050 Β·
IG @iamemily2050 Β·
GitHub @Emily2040
Source intelligence: ByteDance Seedance 2.0 official blog, BytePlus ModelArk docs, Douyin creator community, CSDN practitioner tutorials, April 2026.
Validated repair release focused on reliability and source hygiene.
- Skill frontmatter is normalized for root plus all 23 sub-skills.
- Sub-skills use
metadata.parent: seedance-20. - Current platform/API guidance is stored in
references/api-status.mdand must be rechecked before production advice. - Real-person face, portrait, and voice workflows are authorization-dependent and surface-specific.
- Protected IP, celebrity, brand-logo, studio-style, and exact-scene requests should be routed through
seedance-copyrightbefore final prompt output. - Oversized legacy content is preserved in
references/migrated/and active skills now load lean procedural guidance first.
Validation:
python scripts/validate_skills.py --strict
python scripts/content_audit.py --strict