Vibe Vision Video 0.1.0

The first public release of Vibe Vision Video: build editable video templates once and reuse them across projects and coding agents.

• Portable Agent Skill (skills/vvv), tested in Claude Code and Codex and checked in WorkBuddy; 20+ other agents read the same .agents/skills/ folder (untested). Install as a Claude Code plugin from the repository's marketplace (/plugin install vvv --marketplace actionow-ai/vvv, kit included), with npx skills add actionow-ai/vvv, by copying the directory, or as a ZIP built by scripts/package-skill.py. The helpers find the kit through --kit, $VVV_KIT or the kits/remotion beside the skill.
• Five editable Remotion templates, currently wedding films: Album (16:9 and 9:16), Comedy, Doodle, DV Story and Starter (16:9 and 9:16), plus the reusable papercut-frame effect. Fictional fixtures and checked preview excerpts with music.
• Project helper (vvv.py): templates, describe, init from local photos or the template's demo inputs (--example), validation of every prop against schemas generated from the renderer, samples from the main composition, full renders and stale-output detection. Errors are one JSON object with a stable code.
• Render checks: vvv check measures format, resolution, audio level, still stretches and beat alignment of a render or any video.
• The .vvv package format 1.0 for templates, effects and finished projects, with per-file integrity and license records, previews and render evidence: pack, verify, unpack, add, evidence.
• Reference-video recreation: measure a reference's structure and beat rhythm with the media helper, the music analyzer and vvv check, then rebuild it as an editable project.
• Image generation through a native host tool, Codex's image generation from Claude Code via the bundled c2c bridge, or an explicitly configured Images API, never falling back to a paid route silently.
• Media helper for probing, clipping, audio extraction and frame sampling with FFmpeg.
• Music: a bundled public-domain track, a hash-verified downloader for Mixkit tracks that are not redistributed, and beat analysis for your own music.

Install

• Claude Code: /plugin install vvv --marketplace actionow-ai/vvv
• Codex, Cursor, Gemini CLI, Copilot and 20+ agents: npx skills add actionow-ai/vvv
• WorkBuddy: upload vvv-skill.zip under 技能 → 添加技能
• Any MCP client: https://mcp.vibevision.video/mcp

The .vvv files are ready-made template packages: python3 skills/vvv/scripts/vvv.py verify --package album.vvv, then vvv.py init --template album.vvv --photos ./photos --output ./my-film.

SHA-256

File	SHA-256
album.vvv	5370c88749925833d9d7375d64f36b96a65df5d5492341fc60918c8b78c07723
comedy.vvv	64649f0abcd7fbd3d53005d60886aaf7572972db9a2f0ef758e8a37f316f4ac8
doodle.vvv	b3ef97fcb3efbc72d8f2b94f48474b0cd2f5baaf6206f2d2631702522704c9f4
dv-story.vvv	38224e38c686f892e7750903b6a3993fd0f3bf3ea1641f15362c2e27e35ada94
papercut-frame.vvv	cee615f7fa16301360123e59f432de6ce7f85c145bbeea74b426c91665f839ce
starter.vvv	4b7d65cedb470714d2187a440c62003f0dbcb7a9bd680846f5ad67b71c93a249
vvv-skill.zip	1bce89117e7d8f922edd6f5ff796eeb44f79c775dfdaae1cd148545adcb71dd4