generating-videos
ALWAYS read this skill before generating or animating any video, or calling video_generate — text-to-video, image-to-video, a start→end transition, or a reference / motion / audio-guided clip. Generates a video from a brief — one clip, or several clips joined into a single file at any length. Use when the user wants to make or generate a video, animate a photo, bring an image to life, produce b-roll, film a described scene, or move from one held frame to another. This skill should also be used when the video takes its motion or style from an existing video, or has to run to an existing audio track.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-15
- Stars
- ★ 6
- Forks
- 0
01 前言 Front Matter
- name
- generating-videos
- license
- Apache-2.0
- metadata
- { "summary": "Routes your brief to the best of 10+ SOTA video models (like Veo, Seedance), writes detailed prompts as per model specifications and generates multi-clip stories that hold continuity across cuts, stitched into one file.", "version": "0.5.0", "category": "creative" }
- description
- ALWAYS read this skill before generating or animating any video, or calling video_generate — text-to-video, image-to-video, a start→end transition, or a reference / motion / audio-guided clip. Generates a video from a brief — one clip, or several clips joined into a single file at any length. Use when the user wants to make or generate a video, animate a photo, bring an image to life, produce b-roll, film a described scene, or move from one held frame to another. This skill should also be used when the video takes its motion or style from an existing video, or has to run to an existing audio track.
02 资源文件 1
- 参考文档skill-report.json36.0 KB
03 SKILL.md 正文
Video Generation
Turn a brief into a video.
Workflow
Step 1: Read the brief and the media
Look at everything the user supplied. Run image_analysis on each image and video_analysis on
each video. Write down what each one shows — who is in it, where it is, how it is lit, how it is
framed — so the prompt matches it instead of describing something different. Send a product URL or
photo to analyzing-products. Skip an image you already know: you made it this turn, or you have
already analyzed it.
Then read the brief: how long the video runs, what happens in it, where it will be posted, and whether anyone speaks.
Where the video has a product, write one description of it and reuse it unchanged in every storyboard and every clip prompt — rephrasing it between calls reads as a different product.
- Material and finish, surface by surface.
- Size — its width and height, and how it sits against a hand.
- Two to five visual anchors — features you can verify in the product image and nothing else: exact colours, the shape of a closure or handle, the gauge of a chain or strap, a surface finish, a distinguishing mark.
- What may be done with each part — which parts are fixed to it, and the whole of what the moving ones allow. Nothing later does anything to the product that isn't on this list.
Don't transcribe what the label says. Spelling out the printing invites the model to redraw it, and redrawn text comes back warped.
Step 2: Pick the model and load its parameters
seedance-2.0-fast is the default — the general-purpose workhorse; a weak result is usually the
prompt's fault, not the model's.
The storyboard route always runs on seedance-2.0-fast unless the user named a model. The table below
decides a single clip.
Reach past the default only on a clear signal:
| Reach for another model when the brief… | Model |
|---|---|
| Needs spoken dialogue with lip-sync, the tightest prompt-following, or top cinematic quality | veo-3.1-fast |
| Is editing an existing clip — restyle it, add or replace an element, fix on-screen text — or hinges on legible on-screen text | gemini-omni |
| Must hold one person or character consistent across references or several shots | kling-3.0-pro |
| Has an edgy or sensitive subject Seedance would refuse, or wants the fastest, cheapest draft | grok-imagine-video |
| Is a high-volume or cost-sensitive batch where top quality isn't essential | wan-2.7 |
When more than one fits, work down this list and stop at the first that applies:
- The user named a model — use it.
- The video takes the storyboard route —
seedance-2.0-fast. - Seedance refused the brief (
nsfw/ip) or keeps failing — use the fallback from the table. - The brief matches a row above — use that model.
- Otherwise use
seedance-2.0-fast. If two models tie, use the newer one.
Then call list_video_models for the chosen model and read its accepted modes, aspect ratios, durations, resolutions, and media — each validates its own subset, and you need these before writing the prompt.
Step 3: Route
Take one of three paths:
- The video is a person on camera presenting a product — holding it and talking about it, opening
its package, putting it on, demonstrating it → Trigger the
generating-ugc-videosskill workflow, whether or not the brief says UGC. - One clip covers it — the video runs no longer than the model's longest clip, and holds one
continuous shot → Read and follow the steps in
references/direct-route.md. - It needs more than one clip — the video runs longer than the model's longest clip, or the brief
needs a cut between shots → Read and follow the steps in
references/storyboard-route.md.
list_video_models gives the model's longest clip. Storyboard sheets are what keep a person, a place
and a product looking the same from clip to clip, and what fixes the frame on both sides of a cut, so
anything with more than one clip takes the storyboard route.
A supplied image is used one of two ways. Ask which, if the brief doesn't say:
- As the start frame, the video begins on that exact image.
- As a reference, it shows what a subject or a style looks like and the video is a new scene.
A start frame works for one clip only, and that clip can carry nothing else —
start_frame_image cannot be combined with any reference_* input. Where a person, a product or a
style has to match an image that isn't the start frame, pass that image as a reference and drop the
start frame.
Edge cases
- Safety / NSFW rejection → name a workable stand-in for whatever tripped the filter (cover the
wardrobe, change the setting) and resubmit. If the subject itself is the block, switch to
grok-imagine-video(see Step 2). A second rejection: tell the user which element is blocked instead of retrying blind. - Still generating (
status: "pending") or the call times out → the clip is still rendering on the server, not a failure — do not resubmit (that starts a second billed job). If you got a pending handle, calljob_statuswith it to rejoin; if the call timed out with no handle, wait and tell the user it's still processing rather than firing a fresh generation. - Generic failure (an explicit error result, not a timeout) → read the error. If it names a parameter or a limit, correct that and resubmit. Otherwise resubmit once; if it fails again, either switch to a fallback model (Step 2) and write to its guide, or give the user the error text rather than guessing.
- Reference or frame rejected on count / mode → the error states the model's limit or supported
modes; adjust to it, or check
list_video_models. - Unsure which model fits a named request → call
list_video_modelsand pick bystrengths. - The user wants a standalone audio file, not clip audio — a voiceover to drop on a timeline, or
a music track or sound effect they need to place themselves. Clip audio can't give them that; it is
baked into the picture. Send speech to
generating-audio, and say plainly that separate music and effects tracks aren't generated here. error: "no_provider_configured"→ relay the tool'shint(the user must set their key).
Reference
Each route is a complete workflow. Read the one Step 3 picked, and follow it to the end.
references/direct-route.md— one clip: the six modes and what each one passes, the beats, the prompt and the call.references/storyboard-route.md— more than one clip: the video planned as beats, built through storyboards and joined into one file.
04 同仓库更多 Skills
serpmaven
Scrape Google search results (SERP) to JSON with SERP Maven (serpmaven) - a real Chrome browser on a residential proxy with sticky sessions. Run it as a CLI, a Python library, a local HTTP search API (`serve`), an MCP server or a Docker image; set the proxy login, workers, pages and cost tracking. Use when the user wants Google results, rank tracking, a self-hosted SERP API, or live Google search for an AI agent.
revnu-landing-page-audit
Find out why a landing page isn't turning visitors into signups, demos or sales. Audits one page against what the ad, post or search that sends people there promised, then checks the first screen, proof, call to action, friction and mobile. Returns at most 7 prioritized fixes, plus what is fine and what could not be checked; rewrites specific lines and fixes the page in the repo when asked. Never promises a conversion lift. Works without a Revnu account. Use when the user says or asks: "audit my landing page", "why isn't this page converting", "review my homepage", "my ads get clicks but no signups", "people bounce from the pricing page", "CRO review", "improve conversion rate", "is my hero section clear", "roast my landing page", "rewrite my headline", "what's wrong with this page", "landing page feedback", "make the signup page convert better". Not for SEO or AI-search audits, page speed engineering on its own, or building a new page from scratch.
revnu-growth-context
Write down once who buys, what they're stuck on, what it costs, what proof you have and what you've already tried, so every growth task starts from facts instead of a questionnaire. Creates or updates .agents/revnu-growth.md from the repo, the live site and the user, and imports an existing product-marketing.md. Use when the user says "set up our marketing context", "who is our ideal customer", "write our ICP", "describe the product for marketing", "positioning doc", "what's our brand voice", "remember this about our customers", "log what we tried", "what have we learned from marketing", "we changed pricing, update the context", "update our growth notes", "summarize the business for the agent", "import my product-marketing.md", "start a growth doc", or before a first marketing, SEO, outreach or launch task when no context file exists. Works without a Revnu account.
revnu-get-listed
Get onto the third-party pages AI answers and search cite when buyers ask for the best tool in your category: "best X" roundups, directories, review sites, GitHub awesome lists, community threads and comparison pages. Finds which ones are cited, ranks them by fit and effort, and drafts each submission, PR or pitch by the target's own rules. Use when the user says "get us listed", "which directories should we submit to", "how do we get into best-of lists", "why does ChatGPT recommend competitors and not us", "get mentioned in AI answers", "submit to awesome lists", "get on G2", "where are our competitors listed", "pitch listicle authors", "find backlink opportunities", "launch directories", "get into roundups", "Product Hunt alternatives", "where should we be listed", "get cited by Perplexity". Never submits, pays or posts without a yes. Works without a Revnu account.
revnu-ai-readiness-audit
Audit whether your website gives ChatGPT, Perplexity, Claude, Gemini and Google's AI Overviews something worth quoting, and fix the pages that don't. Use when the user says or asks "is my site ready for AI search", "audit my site for AI answers", "AEO audit", "GEO audit", "why doesn't ChatGPT quote our pricing", "make our pages citable by AI", "do I need llms.txt", "does schema help with ChatGPT", "check our structured data", "is our content visible without JavaScript", "AI SEO audit", "optimize our homepage for AI answers", "LLM SEO check", "why does AI get our product wrong", "fix our site for AI Overviews". Checks fetchability and indexability, answers in the raw HTML, entity facts, schema (including JS-injected markup), headings, dates and llms.txt, then gives at most 10 prioritised fixes and edits the repo when asked. Works without a Revnu account.
proxy-troubleshooting
Diagnose residential proxy failures, with NodeMaven's gateway as the reference - 407 with correct credentials, 406, 410, 500, a connection that hangs, ERR_TUNNEL_CONNECTION_FAILED in a browser, Playwright or Puppeteer ignoring proxy credentials, a sticky session that changes IP, geo-targeting that is silently ignored, passwords with special characters, and unexpected traffic from headless Chrome. Use when a proxy connection fails, returns the wrong location, or costs more traffic than expected.