Seedance prompts

Skill fralapo/awesome-agent-skills/skills/seedance-prompts

Write production-grade prompts for ByteDance Seedance 2.0 video generation. Use when the user wants to generate videos with Seedance 2.0, needs prompt templates for cinematic/anime/commercial/UGC/short-drama/VFX styles, asks about multi-modal references (@图片/@视频/@audio), shot-by-shot timecoded storyboards, camera move vocabulary, character/scene consistency, or video extension/continuity. Triggers on keywords: seedance, seedance 2.0, 即梦, bytedance video, text-to-video prompt.From its SKILL.md

Install

npx -y skills add fralapo/awesome-agent-skills --skill seedance-prompts

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.6 KB, ~1.4k tokens by cl100k_base, as published. Nobody here has run it

Seedance 2.0 Prompt Engineering

Seedance 2.0 is ByteDance's quad-modal video model: image + video + audio + text input. Output: up to 1080p, 4–15s, auto dubbing/scoring. Prompts are high-density storyboards, not one-line descriptions.

When to Use This Skill

User asks to write / improve a Seedance 2.0 prompt
User wants to port a shot idea into Seedance structure
User mentions multi-shot storyboards, camera moves (Hitchcock zoom, FPV, bullet time), character consistency across shots, video-to-video edits, or audio-driven generation
User requests a specific style: cinematic, anime, commercial, UGC, short-drama (竖屏), VFX

Core Prompt Anatomy

Every strong Seedance 2.0 prompt combines 5 blocks. Order is flexible but all blocks matter.

[Style]      → genre, film ref, director ref, render engine, resolution, color grade
[Duration]   → total seconds (4–15s); Seedance caps at 15s
[Scene]      → setting, time of day, weather, atmosphere
[Shot List]  → timecoded blocks ([00-05s], [05-10s], [10-15s]) with shot type,
                camera move, subject action, micro-details, dialogue cue
[Tech/Neg]   → aspect ratio, fps, lens, lighting; negatives (optional 排除: block)

Minimum viable template

Style: {genre/director ref}, {resolution}, {color grade}.
Duration: {N}s.

[00-Ts] Shot 1 ({shot type}): {camera move}. {subject} {action}.
  Details: {micro-detail}. Dialogue Cue: "{line}"
[T-2Ts] Shot 2 ...
[2T-Ns] Shot 3 ...

Tech: {aspect ratio}, {fps}, {lens}, {lighting}.

Multi-Modal Reference Syntax

Seedance accepts explicit reference tags inline. Use sparingly — one ref per asset.

Tag	Meaning	Example
`@图片1` / `@Image 1`	Reference image 1 (identity, style, composition)	`参考@图片1的男人形象`
`@视频1` / `@Video 1`	Reference video (camera move, action, transition)	`完全参考@视频1的所有运镜效果`
`@audio1`	Audio drives rhythm/BGM	`BGM参考@视频3中的音效`

Reference chaining in a single prompt is supported — e.g. identity from image, camera from video, music from audio: 参考@图1的男人形象，他在@图2的电梯中，参考@视频1的运镜.

Three Generation Modes

Mode	When	Key idiom
Text-to-Video	No input asset	Full storyboard + style block
Image-to-Video	Product / character / scene provided	`@图片1` anchors identity; describe motion only
Video-to-Video	Remix / extend / edit	`@视频1` anchors motion or style; describe delta (`将...换成...`)

Category Catalog

Detailed templates and examples for each category live in references/.

references/patterns.md — structural patterns + idioms mined from 195+ prompts
references/templates.md — fill-in templates per category (cinematic, anime, commercial, UGC, short-drama, VFX, sound, consistency, camera, extension)
references/camera-vocab.md — camera moves, lens, lighting, grading vocabulary
references/examples.md — curated working prompts with source attribution

Quick Rules

Always timecode shots. [00-05s] blocks → Seedance's temporal alignment is shot-based, not descriptive.
Three shots is the sweet spot for 15s. Five max. More → drift.
One action per second. Don't pack; over-dense shots desynchronize.
Director/film refs work. Denis Villeneuve style, Wong Kar-wai 90s Hong Kong, Hans Zimmer score → strong tonal anchors.
Dialogue cues inline with lip-sync intent: Dialogue Cue: "line" (200-400ms pauses).
Aspect ratio matters. 9:16 for vertical short-drama / TikTok, 2.35:1 for cinematic, 16:9 default.
Negative prompts use 排除: (Chinese-native pattern) — 排除：模糊，低清，噪点，水印，文字，扭曲变形，五官崩坏.
Character consistency → repeat identity anchor in each shot (同一人物, maintain consistent face/clothing/hair).
For extension/continuity → end shot N with exact state, start shot N+1 from that state (pose, lighting, framing).
Audio drives rhythm — if using @audio, beat-align actions (beat-synced, 与音乐节拍完全吻合).

Anti-Patterns

One-line descriptions → Seedance under-performs vs shot-list format
Mixing Chinese and English mid-sentence without commas → parser trips
More than 2 reference tags per shot → identity conflict
Abstract emotion-only prompts (no visual action) → static output
Over-specifying > 8 micro-details per shot → model drops some

Workflow

Ask user: style, duration, assets on hand (images/videos/audio)
Pick closest template from references/templates.md
Fill 3-shot storyboard with timecodes
Add style header + tech footer + (optional) negative block
If references exist, insert @图片N / @视频N tags
Output prompt in a single fenced block — user pastes into Seedance

What ships with it: 4 files

32.5 KB alongside SKILL.md

references/

Gives 0 of the 12 instructions most video audio skills give in ~1.4k tokens

Counted across 619 of the 725 authors here whose files we hold, read 2026-09-06

Read product marketing context firstin 13 of 619, across 7 files
Define the core visual thesis in one sentencein 11 of 619, across 3 files
Break the concept into 3 to 6 scenesin 11 of 619, across 3 files
Render the smallest working version firstin 11 of 619, across 3 files
Start with a low-quality smoke test renderin 11 of 619, across 3 files
Add captions for accessibility and engagementin 11 of 619, across 5 files
Write the scene outline before writing codein 11 of 619, across 3 files
Specify subject, action, camera, style, and moodin 11 of 619, across 5 files
Decide what each scene provesin 10 of 619, across 2 files
Export one clean thumbnail framein 10 of 619, across 2 files
Pick the right tool for the jobin 10 of 619, across 4 files
Run the test suite before proposing a fixin 8 of 619, across 7 files

Said here and by no other author read

Always timecode shots using temporal alignment blocks
Limit to three shots for fifteen seconds
Include one action per second maximum
Use director and film references as tonal anchors
Place dialogue cues inline with lip-sync intent
Specify aspect ratio in the tech footer

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.