app.video-studio

@lssm-tech/app.video-studio

Thin Remotion Studio wrapper. Re-exports the Remotion entry point from @lssm-tech/lib.video-gen.

What It Does

• Couche: app (Remotion Studio)
• Consumers: internal — video preview and rendering

Running Locally

From packages/apps/video-studio:

• bun run dev
• bun run build

Usage

bun run dev

Local Commands

• bun run dev — bunx remotionb studio src/index.ts
• bun run build — echo 'No build step needed for Remotion Studio'
• bun run clean — rimraf out .turbo
• bun run test — standalone host test: imports the Remotion entry and lists its compositions through Remotion's composition manager in happy-dom (no frame render)
• bun run render — bunx remotionb render src/index.ts
• bun run render:all — bunx remotionb render src/index.ts ApiOverview out/api-overview.mp4 && bunx remotionb render src/index.ts SocialClip out/social-clip.mp4 && bunx remotionb render src/index.ts SocialClipSquare out/social-clip-square.mp4 && bunx remotionb render src/index.ts SocialClipPortrait out/social-clip-portrait.mp4 && bunx remotionb render src/index.ts TerminalDemo out/terminal-demo.mp4

Recent Updates

• Replace eslint+prettier by biomejs to optimize speed
• Improve agentpacks
• Switch Remotion CLI to bunx remotionb for Bun runtime support

CompanyOS film workflow

Three ordered authoring-time scripts produce a locale-matched render for the "Room for the real work" campaign:

1. bun run companyos:narrate — synthesizes narration through ElevenLabs once per content-hash and caches audio under public/companyos/narration/. Writes a manifest.json recording each take's locale, key, text, hash, duration, and voice cast. ELEVENLABS_API_KEY is required only for lines not yet cached; cached lines are replayed for free. Narration is generated once at authoring time. Never call this at build time or page load.
2. bun run companyos:mix — reads the narration manifest, builds each locale's final WAV mix (narration + procedural sound design + music bed with ducking), normalizes loudness, and writes render props to public/companyos/mix/owners-first-article-{locale}.props.json. The props JSON carries the audio path and the computed OwnersFirstArticleTimeline derived from measured narration durations.
3. bun run companyos:render [locale…] — renders CompanyOsOwnersFirstArticle-{locale} via bunx remotionb for each requested locale (all three if none specified), reading props from the mix output. Outputs go to out/companyos/.

Run the scripts in order: narrate → mix → render. The mix script depends on the narration manifest; the render script depends on the mix props.

Voice casting is recorded in scripts/companyos/narrate.ts (VOICE constant). Change the voiceId there to recast; the hash changes and the new lines are synthesized on the next narrate run.

Notes

• Do NOT add compositions here. All compositions live in @lssm-tech/lib.video-gen.
• This package exists solely to provide a Remotion Studio entry point and authoring-time scripts.
• Uses remotionb CLI for Bun runtime support. Known caveat: SSR scripts may not auto-quit (see Remotion Bun docs).
• Narration is generated once at authoring time from cached content hashes. Never call companyos:narrate at build or page load. ELEVENLABS_API_KEY is only required for uncached lines.