actionow-ai/vvv

[!NOTE] For learning and research only. This project is for learning and researching reusable video templates, beat-based editing and agent workflows. The couples and photos in the examples are fictional. Recreating a reference film is for studying its structure and rhythm; do not repost or pass off someone else's work, and use only photos, music and fonts you have the right to use.

Made with the templates

Doodle · hand-drawn	Album · clean album	DV Story · camcorder
Comedy · comic opener	Starter · beat-cut reel	Album 9:16 · vertical short
Editorial · clean magazine	Cinema · venue-screen opener	Memoir · handwritten life story

_{Click a loop for the full preview with sound. All rendered straight from the templates and passing vvv check; the couples and photos are fictional. Newest: Editorial, Cinema and Memoir, recreated from reference films: a clean magazine-style MV, an opener on black for the venue screen, and a handwritten life story on film. All templates, formats and styles}

From a folder of photos to a finished film

Hand your agent the photos and the idea; from sorting the photos to rendering the film, it can take every step. You say what you want, watch the sample and decide.

Photos	The idea	Sample and revisions	Final film and keeping it
01 Idea Tell the agent the style, format and song you want, in a sentence or two.	04 Brief Style, format, length and song pinned down. A song you name comes from a file you have, or from musicdl if you installed it (personal use only).	07 Sample The music is fitted so the chorus opens the climax page; you hear the length and seconds per page before anything renders.	10 Final render A 1080p film, checked for format, loudness, still frames and beat sync.
02 Photos A local folder, or a 百度网盘 share link.	05 Photo read Each photo tagged by who, mood and outfit; one outfit per page, and paired pages change outfit.	08 Notes Watch the sample and say what to change.	11 To your drive Save the film to your 百度网盘 when you want, only when you say so.
03 Download The agent downloads the originals and scales copies to render size; the originals stay untouched.	06 Treatment One of three: the best-fitting template, a template you pick, or a film you like, studied for its structure and rhythm and rebuilt with your own material (抖音, 小红书, 快手 or TikTok links work directly).	09 Revise Photos swapped, copy rewritten, pacing adjusted, one note at a time; only the changed range re-renders.	12 Keep the project The editable project is kept, so you can change it later or hand it to another agent.

Use the vvv skill to make a wedding MV: the photos are in 百度网盘 at https://pan.baidu.com/s/… (code abcd), the song is 咱们结婚吧, and the style should follow this 小红书 video https://xhslink.cn/o/…. The couple are Lin Yizhou and Xu Zhixia, married on October 18. Show me a sample first.

Add it to your agent

Any one of:

Agent	Install
Claude Code	/plugin install vvv --marketplace actionow-ai/vvv (the plugin includes the templates)
Codex, Cursor, Gemini CLI, Copilot and 20+ more	npx skills add actionow-ai/vvv, then clone this repository and set VVV_KIT (details)
WorkBuddy	Upload the skill ZIP from Releases under 技能 → 添加技能 (details)
Any MCP client	claude mcp add --transport http vvv https://mcp.vibevision.video/mcp to browse templates and guides with no install

Rendering runs on your machine, which needs Python 3.9+, Node.js 26 and npm; the agent installs everything else. Optional, under your own accounts: a TikHub token for share-link references, a 百度网盘 login, and musicdl (integrations).

One template, every new film

The scenes, layouts, timing and checks are built once. A new film changes only what belongs to it.

Built once	Scene structure and camera moves · layouts for 16:9 and 9:16 · beat-planned timing · render checks
Changed per film	Photos · names and dates · every line of copy · music and format

vvv.py describe tells the agent each template's photo slots, copy fields and settings, and validate catches bad inputs before a render. The agent changes the template's inputs first and its React/TypeScript source only where the brief needs it. Reuse is meant to save the tokens otherwise spent redesigning, regenerating and debugging the same video code; no comparative benchmark is published yet (see the roadmap).

Show it a film you love. Get your own.

Give your agent a reference video and the attachments that make it yours: photos, a logo, copy, music. It studies how the reference works and rebuilds it as an editable project with your material.

01 Reference + attachments A video file or a 抖音, 小红书, 快手 or TikTok share link, plus your photos (local or from 百度网盘), logo, copy and music. The reference is studied, never shipped.

02 Inventory Every page in beats: frames every 0.1 s around each transition, photo boxes measured from settled frames, how each element enters and leaves. Written down before any code.

03 Rebuild and compare On the closest template or a new composition. Missing artwork is generated from reference crops; frames are compared side by side and every off-beat peak is listed.

Use the vvv skill to recreate this 小红书 video https://xhslink.cn/o/… as our own film with the photos in ./photos, our logo and copy. Generate any missing artwork and show me a checked sample before the full render.

Best with an image-capable agent. With Claude Code and Codex both installed, Claude Code can use Codex's image generation through the bundled bridge. Video and voice are not generated. How it works

Also included

• Photos in emotional order: every template maps its emotional sections (laughs before tears, first date to wedding). The agent looks at each photo, tags who is in it, its mood and when it was taken, places it in the matching section, and shows you the order before rendering.
• Style presets: settle the opening rule, the music and the credit line once, save them, and every later film starts from them.
• Render checks: vvv check measures format, audio, still stretches and beat alignment, names the peaks furthest off the beat, then lists what only a viewer can judge.
• Template updates: vvv.py update brings an existing project's template code up to the kit's current version and keeps your own edits.
• Handoff: pack a template, effect or finished project as one .vvv file with its code, media licenses and render evidence, so any agent can continue it.
• Footage tools: a local FFmpeg helper probes media, cuts ranges, extracts audio and samples frames (walkthrough).
• Community templates: built one you are proud of? Pack it and submit it on the website. Once approved, it appears on the site and in the MCP server.

The first time your agent runs commands on your machine, it may ask for permission. To run things by hand or build a new template, see getting started and contributing; ready-made .vvv template packages are attached to Releases.

Status 0.2, an early release. Verified end to end on macOS ARM64 with Claude Code and Codex; WorkBuddy and other platforms are less tested. Verification and limits

Use and license For learning and research only. Code, docs, templates and demo media are MIT; fonts, preview soundtracks and third-party code keep their own terms, and musicdl is non-commercial only: media · third-party notices

Get involved Contributing · Roadmap · Changelog Security · Brand kit · Submit a template

_{© 2026 Vibe Vision Video contributors · vibevision.video}