Zen Foley

Video-to-audio Foley: high-fidelity sound-effect generation synchronized to video.

Status: work in progress. Built on a permissive, commercially-licensed foundation (no non-commercial or restricted-weight dependencies). The Zen model is not yet trained/integrated here — this provides the licensing, attribution, and integration scaffold.

Foundation

Built on MMAudio (https://github.com/hkchengrex/MMAudio, MIT) — a permissive, commercial-friendly video-to-audio foundation.

License

MIT License — see LICENSE and NOTICE. Copyright 2025-2026 Zen Authors (https://zenlm.org).

Roadmap

• Wire the MMAudio backbone as the V2A pipeline
• Re-train on commercially-cleared foley/audio data
• Inference, evaluation, and release