跳到正文
cccskills
免费在 GitHub 公开

music-video

Turn a song into a beat-synced music video with Guaardvark's Director: song analysis, per-cut shot prompts, storyboard keyframes, image-to-video renders, cuts on the beat. Use when the user says "music video", "video for this track", or wants visuals timed to a song. Has an approval gate before GPU spend.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(1)

  • SKILL.md4.2 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Music video with Guaardvark

Read setup first. Needs the comfyui plugin and an installed i2v model (wan22-5b by default; minimax-h3-int8 for frame-matched continuity).

Start: MCP generate_music_video

  • song: document id or a filesystem path to mp3/wav/flac/ogg. A path is uploaded for the user.
  • style_prompt: mood, palette, movement, era ("grainy 16 mm, neon rain, slow dolly, 1984").
  • name (defaults to the song file name), i2v_model (defaults to the active i2v model).
  • The tool analyses the song (tempo, beats, sections), writes one unique prompt per cut and stops at the approval gate. It does not spend GPU rendering clips. Tell the user that.

The gate and the rest of the pipeline (REST)

B=${GUAARDVARK_URL:-http://localhost:5000}, ID = the project id the tool returned.

stepcallwhat it does
read the planGET $B/api/music-video/$IDcuts, prompts, status, storyboard state
edit promptsPOST $B/api/music-video/$ID/plan with the edited cut prompts (and optional global style)operator edits before render
regenerate the planPOST $B/api/music-video/$ID/regenerate-plan or /replannew prompts from the Director
storyboards firstPOST $B/api/music-video/$ID/generate-storyboardsone keyframe still per cut, thumbnails-first review
redo one keyframePOST $B/api/music-video/$ID/regen-storyboard/<idx>keeps the others
approvePOST $B/api/music-video/$ID/approvereleases per-clip generation. Only on the user's explicit yes
re-analysePOST $B/api/music-video/$ID/analyzerestart song analysis
cancelPOST $B/api/music-video/$ID/cancel
list allGET $B/api/music-video (?project_id=<id> or ?project_id=none to filter)
progressGET $B/api/music-video/$ID/progressstages, the clip in flight (sampler step/steps), ETA. Report this instead of guessing
rename / file under a projectPATCH $B/api/music-video/$ID with {name?, project_id?, notes?, tags?}the finished video joins that project's Documents
edit one cutPOST $B/api/music-video/$ID/cut/<idx> with {prompt?, seed?, locked?}at the approval gate it only changes the plan; while generating it re-renders that cut: ask first
reopen a finished videoPOST $B/api/music-video/$ID/reviseback to the approval gate, rendered clips kept; on approve only changed cuts render. /replan re-renders all
copyPOST $B/api/music-video/$ID/duplicate with {name?, project_id?, keep_clips?}new video at the approval gate
versionsGET/POST $B/api/music-video/$ID/revisions, POST .../revisions/<rid>/restoresaved on approve, reassemble, revise, restore; restore lands at the approval gate
a project's videosGET $B/api/projects/<id>/videosmusic videos and Film Crew filed under it

Creating a project without the MCP tool: POST $B/api/music-video with {"name", "song_document_id", "style_prompt", "settings": {"i2v_model": "..."}, "user_treatment": "optional short story", "project_id": optional}. The MCP tool takes project (a Project id or exact name). user_treatment becomes the creative source the Director writes cuts from.

How to run it well

  1. Start the project, show the cut list and the beat count.
  2. Offer storyboards before approval: they are cheap and catch a wrong style early.
  3. Ask for approval in plain words with the cost: number of cuts times the clip time for the chosen model. Approve only after a clear yes, and only once GET $B/api/music-video/$ID shows current_stage awaiting_approval with cuts: before analysis finishes the approve route answers 409.
  4. Character in the video: create a Cast subject and train a LoRA first (the cast skill) so the Director can lock identity across cuts.
  5. The finished cut is assembled to the detected beat grid. The Video Editor page can trim or overlay text afterwards.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

cast

免费

Build consistent characters, environments and props in Guaardvark's Cast Library and train LoRAs for them locally (reference photos → vision bible → sample plan → approved samples → training). Use when the user wants the same face or object across images, videos, a music video, or a Film Crew production, or asks to "train a LoRA".

中文简介正在准备中。当前显示原文说明。

guaardvark/guaardvark2732026年10月22日 更新

code

免费

Code intelligence over repositories Guaardvark has indexed: semantic code search, exact-text search, file listing, PageRank repository map, import dependency graph, reading one class or function, analysis and generated rewrites, plus the System Mapper and self-improvement status. Use when the user asks where something is implemented, how modules depend on each other, or wants an architecture map of an indexed repo.

中文简介正在准备中。当前显示原文说明。

guaardvark/guaardvark2732026年10月22日 更新

film-crew

免费

Run Guaardvark's five-role Film Crew locally: screenwriter, casting (Cast Library LoRAs), cinematographer storyboards, keyframe renders, editor. Turns a logline or screenplay into a finished short video with human gates at casting and storyboard. Use when the user says "film this", "make a short film", "storyboard this script", or "start the film crew".

中文简介正在准备中。当前显示原文说明。

guaardvark/guaardvark2732026年10月22日 更新

image

免费

Generate or edit images on the user's own GPU through Guaardvark: single images, instruction edits, background cut-outs, inpaint and outpaint, consistent characters from the Cast Library, and batch runs of many prompts. Use when the user asks to create, draw, render, visualize, edit, or batch-generate images locally.

中文简介正在准备中。当前显示原文说明。

guaardvark/guaardvark2732026年10月22日 更新

knowledge

免费

Answer from the user's own indexed documents with Guaardvark's local RAG, browse and read indexed files section by section, remember facts across sessions, and process or fetch new material. Use when the user asks what their documents say, wants a source-cited answer from their files, says "remember this", or asks to index a PDF, DOCX, spreadsheet, image or URL.

中文简介正在准备中。当前显示原文说明。

guaardvark/guaardvark2732026年10月22日 更新

models

免费

Add any Hugging Face image or video model, checkpoint or LoRA to Guaardvark from a URL, list what is installed, and download registry models on request. Use when the user pastes a huggingface.co link or a .safetensors URL, asks "can it run <model>", or wants a new LoRA or checkpoint available in the Studio.

中文简介正在准备中。当前显示原文说明。

guaardvark/guaardvark2732026年10月22日 更新

查看 guaardvark 的全部技能

报告此技能的问题