anycap-seedance-2
Compile user intent, scene beats, and reference assets into compact, controllable Seedance 2.0 prompts executed through the AnyCap CLI. Use when creating, testing, iterating, troubleshooting, or continuing Seedance 2 / Seedance 2 Fast video generations with AnyCap, especially when prompt structure, reference consistency, image/video/audio role mapping, product or character identity, multi-modal references, continuation, retakes, or generated-video review matter.
- Installs
- 43
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/anycap-ai/anycap --skill anycap-seedance-2- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads anycap-seedance-2 automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
AnyCap Seedance 2
Overview
Use this skill as a Seedance prompt compiler and reference consistency controller for anycap video generate. Its job is to turn user intent, scene beats, and reference files into a compact prompt contract that Seedance can follow without drifting, overloading the clip, or contaminating one reference with another.
Requires the AnyCap CLI binary, internet access, and AnyCap authentication.
For command details, read skills/anycap-cli/SKILL.md and skills/anycap-cli/references/video-generation.md. Read references/reference-consistency.md when references, identity, continuity, or role conflicts matter. Read references/workflow.md for review/retake logic. Read references/patterns.md when the user asks for an ad, product shot, UGC-style clip, character shot, transformation, motion-transfer, or continuation pattern.
Operating Promise
Do not make the user become a prompt engineer. Convert their request into a prompt contract. If a default is reasonable, choose it and say what you chose. Ask only when the missing choice changes cost, safety, rights, aspect ratio, reference ownership, or the core creative direction.
Default to this low-friction path for a first draft:
seedance-2-fast, text-to-video or image-to-video, 4 seconds, 16:9, 720p, generate_audio=false
After the draft lands, review whether the prompt contract held: beat, endpoint, reference identity, allowed motion, exclusions, and continuity. Offer the smallest next move: keep, final render, one-variable retake, continuation, or post-production fix.
Quick Defaults
- Use
seedance-2-fastfor drafts, previewing, and low-friction user tests. - Use
seedance-2for final-quality takes, higher resolution, or when the user explicitly asks for best quality. - Default draft settings:
duration=4,aspect_ratio=16:9,resolution=720p,generate_audio=false. - For final standard takes, prefer
resolution=1080punless the schema or user target says otherwise. - Always run
anycap video models <model> schema --mode <mode>before generation and follow the returned schema. - Always use
-owith a descriptive filename. - Run
anycap statusfirst if authentication or connectivity is uncertain.
Fast Routing
Choose the production path before writing prompt prose:
| User intent | Path |
|---|---|
| One simple idea, no references | Fast clip: text-to-video |
| Animate an uploaded/generated still | Image animation: image-to-video |
| Use product/identity plus motion/audio references | Reference-led clip: multi-modal-reference |
| Make another part from an accepted take | Continuation: multi-modal-reference with previous clip |
| Ad, UGC, product demo, transformation, character intro | Load references/patterns.md and choose a pattern |
| Bad or partial result came back | Load references/workflow.md and use Review/Retake |
| Reference role is ambiguous or several assets compete | Load references/reference-consistency.md |
Mode Choice
- No references: use
text-to-video. - Image reference only: use
image-to-videoand pass--param images=<path-or-url>. - Image plus video or audio references: use
multi-modal-reference. - Continuation from an accepted clip: use
multi-modal-referencewith the accepted clip asvideos=<path>. - If the user wants to animate a still image, keep the prompt focused on motion, camera, light, and endpoint instead of re-describing everything in the image.
Prompt Compiler
Before writing the final prompt, form a short internal brief. Do not show this unless useful:
- Reference contract: exact tags, primary roles, allowed transfer, exclusions.
- Subject lock: what identity/product/character must remain stable.
- Clip beat: one visible action that can happen inside the chosen duration.
- Endpoint: what state the clip stops on.
- Camera: one motivated movement.
- Light: one physical source or atmospheric change.
- Sound: ambience, silence, SFX, music energy, or
generate_audio=false. - Continuity constraints: what must not change, transfer, appear, replay, or happen yet.
Then send only compact natural language to Seedance. Avoid generic boosters like "ultra cinematic" unless the user asked for that exact style. Keep the final prompt around 40-110 words for ordinary clips. When references exist, the reference contract comes first.
Use this final prompt shape:
[Reference contract]. [Subject lock]. [One visible action]. Camera: [one move]. Light: [physical source/change]. Sound: [audio intent or silence]. Stop when [endpoint]. Constraints: [do not change/transfer/replay].
Translate abstract requests into visible direction:
| Abstract ask | Compile as |
|---|---|
| cozy | warm practical light, small hand motion, slower camera |
| premium | controlled product highlight, still endpoint, clean background |
| tense | held frame, delayed movement, narrower light, quieter sound |
| energetic | faster blocking, rhythmic cuts or push, brighter contrast |
| realistic | physical cause for motion, restrained camera, ordinary light |
Reference Consistency Contract
Give every reference a single primary role and an explicit exclusion. Treat this as a contract, not decoration.
- Image: identity, product, pose, costume, environment, first frame, or last frame.
- Video: motion, camera rhythm, pacing, blocking, gesture timing, or accepted continuation source.
- Audio: tempo, mood, ambience, music texture, or delivery tone.
Write both transfer and exclusion clauses:
[Image1] controls product identity. [Video1] controls camera pace only. Preserve the product from [Image1]; do not copy people, logos, room, or costume from [Video1].
Do not let a reference carry everything. If a user says "use this as style," decide whether they mean identity, environment, color, motion, or pacing, then state that role in the prompt.
Use these conflict rules:
- Identity beats style: product/character identity reference wins over mood/style references.
- Accepted footage beats plan: a generated accepted clip defines the next opening state.
- Motion references transfer timing and blocking only, not people, rooms, logos, costumes, or faces.
- Audio references transfer tempo/energy only unless the user has explicit rights for voice/song likeness.
- Exact text, logos, captions, and UI labels belong in source images or post-production, not newly invented by video generation.
Quality Gates
Before generating, check:
- One visible beat fits the duration.
- Endpoint is explicit.
- Camera has one main move, not a list.
- Light has a physical source or transition.
- References have roles and exclusions.
- Reference tags are preserved exactly from prompt to command.
- Text/logos/subtitles are not expected to be generated perfectly inside the video.
- Any real person, brand, voice, song, or protected IP has been rewritten or confirmed as authorized.
Commands
Text-to-video draft:
anycap video generate \
--model seedance-2-fast \
--mode text-to-video \
--prompt "A compact scene prompt..." \
--param duration=4 \
--param aspect_ratio=16:9 \
--param resolution=720p \
--param generate_audio=false \
-o seedance-scene-draft.mp4
Image-to-video:
anycap video generate \
--model seedance-2-fast \
--mode image-to-video \
--prompt "[Image1] preserved; the drawn steam becomes soft moving vapor while the cup, window, table, and composition stay fixed. Camera: slow push-in. A warm morning highlight crosses the cup. Stop on the brightest highlight. Do not add text." \
--param images=./reference.png \
--param duration=4 \
--param aspect_ratio=16:9 \
--param resolution=720p \
-o seedance-image-draft.mp4
Multimodal reference:
anycap video generate \
--model seedance-2 \
--mode multi-modal-reference \
--prompt "[Image1] controls subject identity; [Video1] controls camera rhythm only; [Audio1] controls tempo only. Do not transfer people, logos, rooms, voices, or music identity from the references." \
--param images='["./subject.png"]' \
--param videos='["./motion-ref.mp4"]' \
--param audios='["./tempo-ref.mp3"]' \
--param duration=8 \
--param aspect_ratio=16:9 \
--param resolution=1080p \
-o seedance-multimodal-final.mp4
Review And Retake
After generation, inspect the file and, when useful, run:
anycap actions video-read --file ./seedance-scene-draft.mp4 --instruction "Describe the visible action, camera movement, lighting, audio, and final frame. Note mismatches against the prompt."
Classify the take as keep, fix in post, edit, re-roll, or rewrite. For retakes, change one variable at a time: prompt clause, mode, reference, duration, resolution, or model. Use seedance-2-fast for exploration and reserve seedance-2 for the locked direction.
Return results in this compact form:
Done: <local_path>
Model/mode: <model> / <mode>
Params: <duration>, <aspect_ratio>, <resolution>, audio <on/off>
Take: <one-sentence review>
Next: <keep/final render/retake/continue/post>
FAQ
What does the anycap-seedance-2 skill do?
Compile user intent, scene beats, and reference assets into compact, controllable Seedance 2.0 prompts executed through the AnyCap CLI. Use when creating, testing, iterating, troubleshooting, or continuing Seedance 2 / Seedance 2 Fast video generations with AnyCap, especially when prompt structure, reference consistency, image/video/audio role mapping, product or character identity, multi-modal references, continuation, retakes, or generated-video review matter.
How do I install anycap-seedance-2?
Run npx skills add https://github.com/anycap-ai/anycap --skill anycap-seedance-2 in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/anycap-seedance-2/ (personal) or .claude/skills/anycap-seedance-2/ (project).
Which AI agents can use anycap-seedance-2?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is anycap-seedance-2 safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains anycap-seedance-2?
It is published by anycap-ai in anycap-ai/anycap. It has 43 recorded installs.
More from anycap-ai/anycap
anycap-media-production
anycap-ai/anycap
Produce media assets using AnyCap: generate images, videos, music, speech, dialogue, and complete audio scenes from text or reference inputs, refine images through interactive visual annotation, and deliver finished assets. Covers the full production workflow from concept to delivery across all media types (image, video, music, audio). Use when creating images, videos, music, voice content, dialogue, complete audio scenes, or any visual/audio content -- including iterative refinement with human feedback. Also use for image-to-image transformation, video generation from images, audio generation from references, and annotation-driven precise edits. Trigger on: media production, asset generation, generate image/video/music/audio, create visual content, produce assets, iterative image editing, annotate and refine, creative workflow, content creation, or any task requiring AI-generated media output.
anycap-deepresearch
anycap-ai/anycap
Guide for conducting thorough, multi-source research and producing comprehensive, well-sourced reports. Powered by AnyCap -- the capability runtime that equips AI agents with web search (including AI Grounded citations), web crawl, image generation, cloud storage, and one-click web publishing through a single CLI. Use when the user asks for deep research, competitive analysis, market research, technical deep dive, literature review, technology comparison, or any task requiring multi-source information gathering and synthesis. Also use when users say "investigate", "survey the landscape", "compare X vs Y", "state of the art", "write a report on", "look into", "find out about", "analyze the market", or any inquiry that needs more than a single search. Trigger on mentions of research, analysis, investigation, comparison, report, survey, or deep dive.
anycap-ai-tool-seo
anycap-ai/anycap
Audit and plan organic search for AI tool, SaaS, and product-led websites. Use for Search Console diagnosis, existing-page optimization, search intent and page ownership, technical indexing, content briefs, international SEO, citations, or programmatic SEO. Start with measured visibility and existing URLs before recommending new pages.
anycap-human-interaction
anycap-ai/anycap
Collect structured visual feedback from humans using AnyCap's annotation tool, or create and iterate on diagrams using the interactive whiteboard (Excalidraw). Covers image annotation, URL/web page review with screen recording, video review, audio feedback, and collaborative diagramming with Mermaid input. Use when you need a human to point at things, mark regions, draw on screenshots, review a web page or UI, narrate feedback over a recording, provide any spatially-grounded visual input, create or iterate on architecture diagrams, flowcharts, or wireframes. Also use when you need to present work-in-progress to a human for approval or revision. Trigger on: get feedback, show to user, review UI, annotate, mark up, visual feedback, screen recording, user review, human-in-the-loop, approval flow, interactive review, whiteboard, diagram, draw, flowchart, wireframe, or architecture chart.
anycap-social-meme-workflows
anycap-ai/anycap
Create meme-style images, funny meme drawings, captioned photos, and lightweight social visuals with AnyCap. Use when an agent needs to turn a joke, reaction, campaign hook, quote, or screenshot into a meme image, funny doodle-style meme drawing, captioned image, or short meme-video concept. Combines AnyCap image generation and editing with deterministic local text overlay, optional video generation, and Drive or Page delivery.
anycap-blog-production
anycap-ai/anycap
Turn user-provided data, notes, research, product facts, or structured inputs into AnyCap-style blog posts and then enrich those posts with first-party evidence blocks when needed. Use when a user wants a blog, article, tutorial, learn page, or SEO content draft written in the tone of the AnyCap website. Best fit when the input is data-first rather than prose-first: spreadsheets, JSON, bullets, notes, source URLs, benchmarks, product facts, or rough research. Trigger on: write blog from data, turn notes into article, blog drafting, article production, learn page drafting, SEO article in AnyCap tone, or requests to convert structured inputs into publish-ready content.
Popular media & creative skills
hyperframes
heygen-com/hyperframes
Read first to make, edit, animate or render any video with HyperFrames (HTML rendered to video), including launch films, explainers, music videos, charts, maps, vertical clips, title cards, captions and overlays. Holds the render contract, the work loop and what each kind of video needs.
media-use
heygen-com/hyperframes
Get or make any media for a HyperFrames video - music, sound effects, images, icons, logos, voiceover, avatar and generated video, transcripts, cutouts, colour grades and looks - with HeyGen or free local tools, and mix placed audio. Also for footage that looks dark, flat or dull, or should be stylized or hidden.
hyperframes-registry
heygen-com/hyperframes
About 450 ready-made HyperFrames blocks and components installed as editable source - transitions, overlays, titles, kinetic type, captions, charts, map overlays, 3D, camera moves. Use it before hand-building any named visual (CRT scanlines, glitch, film grain, a chart); also for hyperframes add and catalog.
hyperframes-creative
heygen-com/hyperframes
Non-animation creative direction for HyperFrames videos. Use for design spec (frame.md / design.md) handling, palettes, typography, narration, beat planning, audio-reactive visuals, composition patterns, and brand / style decisions. For atomic motion patterns and scene blueprints, use hyperframes-animation.
ai-video-generation
101-skills/superpowers
Generate AI videos with Google Veo, Seedance 2.0, HappyHorse, Wan, Grok and 40+ models via inference.sh CLI. Models: Veo 3.1, Veo 3, Seedance 2.0, HappyHorse 1.0, Wan 2.5, Grok Imagine Video, OmniHuman, Fabric, HunyuanVideo. Capabilities: text-to-video, image-to-video, reference-to-video, video editing, lipsync, avatar animation, video upscaling, foley sound. Use for: social media videos, marketing content, explainer videos, product demos, AI avatars. Triggers: video generation, ai video, text to video, image to video, veo, animate image, video from image, ai animation, video generator, generate video, t2v, i2v, ai video maker, create video with ai, runway alternative, pika alternative, sora alternative, kling alternative, seedance, happyhorse
ai-image-generation
101-skills/superpowers
Generate AI images with GPT-Image-2, FLUX, Gemini, Grok, Seedream, Reve and 50+ models via inference.sh CLI. Models: GPT-Image-2, FLUX Dev LoRA, FLUX.2 Klein LoRA, Gemini 3 Pro Image, Grok Imagine, Seedream 4.5, Reve, ImagineArt. Capabilities: text-to-image, image-to-image, inpainting, LoRA, image editing, upscaling, text rendering. Use for: AI art, product mockups, concept art, social media graphics, marketing visuals, illustrations. Triggers: flux, image generation, ai image, text to image, stable diffusion, generate image, ai art, midjourney alternative, dall-e alternative, text2img, t2i, image generator, ai picture, create image with ai, generative ai, ai illustration, grok image, gemini image, gpt image, openai image, chatgpt image