Kompo.ai — Service Overview
Kompo.ai is an LLM-guided music video composition platform. You describe what you want in
natural language, upload source files, and the platform composes, renders, and delivers a
finished music video.
What you can do with kompo.ai
The API surface supports six workflow flows:
- Bootstrap & discovery — the tools manifest describes every callable
operation, its HTTP method, path, and input schema.
- Authentication — request an account (invitation-only), then log in via
kli auth/url
→ kli auth/complete <callback-url> (PKCE paste-back flow). The CLI auto-refreshes
tokens via kli auth/refresh.
- Upload — upload audio, video, and image files to your media library via presigned
S3 URLs.
- Analyze & tag kilder — create a kilde (a metadata-tagged, reusable media source
with segment definitions) that references your uploaded files. Kilder are the building
blocks kompositions reference. Tagging here means capturing the track’s beat grid
and giving its structural parts human-meaningful names — the names the user will
compose with. See source-metadata-approach.
- Compose a komposition — create a markdown document (a komposition) that describes
your music video: BPM, beat timings, visual tracks, and which kilder to pull from.
You can compose via natural-language chat or the CRUD API directly. The user describes
structure in bars and beats and the system does the timing arithmetic — see
beats-and-bars.
- Build, poll & download — submit the komposition for rendering. Poll the job status
until it succeeds, then download the finished
.mp4 from the outputs endpoint.
Environments
The production frontend and API live at https://ai.makeshitapp.com; the API is path-routed
under /api/*.
Quick health check
curl https://ai.makeshitapp.com/api/health
# → { "status": "healthy", "version": "2.0-typescript-full", "timestamp": "..." }
No authentication required.