Facets

Facets is a multimodal listening instrument that turns locally measured audio features into an image-generation brief, sends that brief through the Livepeer Agent API, and preserves a visible receipt of what was measured, directed, and generated.

Raw audio stays in the browser. Only labelled measures and the user's creative direction are sent to Livepeer Agent.

Try the public app · Watch the 39-second demo

Hackathon track

Livepeer Agent Builder

Livepeer Agent is central to the working loop:

1. The browser measures pulse, density, brightness, and drift from a chosen audio file.
2. A person adds or edits a creative direction.
3. Facets quotes flux-schnell pricing through the required Livepeer Creative MCP, then calls create_media with a $0.02 per-image cap.
4. The generated image returns to the instrument with a provenance-style generation receipt.
5. The person can revise the direction and generate another facet.

Run locally

Facets is a dependency-free static application.

python -m http.server 8000 -d dist

Then open http://127.0.0.1:8000.

Demo

The 39-second MP4 walkthrough shows calibration measures, a Livepeer Creative MCP image-generation call with a price quote, and a second call refining the same sound into a different visual direction. A WebM copy is also included in demo/.

Privacy and scope

• Audio analysis happens locally in the browser using the Web Audio API.
• Raw audio is not uploaded by Facets.
• The generated visual request contains labelled measurements and human direction.
• This hackathon build focuses on the complete sound-to-visual loop. The future Elsehow return path would translate the visual result back into spatial sound.

Built by

Miranda Price / Forgewrld711, with Morrow.