---
title: MEGAMI
emoji: 🎚
colorFrom: yellow
colorTo: gray
sdk: gradio
sdk_version: 5.28.0
python_version: "3.10"
app_file: app.py
pinned: false
license: cc-by-nc-sa-4.0
short_description: Automatic Mixing via Generative Model of Effect Embeddings
---

# MEGAMI — Automatic Music Mixing

Upload 2–6 **dry (unprocessed) mono/stereo instrument stems** (at least ~12 s each) and MEGAMI will generate a professionally mixed stereo output.

**Requirements:**
- Each track must be at least 11.9 s long (525 312 samples @ 44 100 Hz)
- Supported formats: WAV, FLAC, MP3 (auto-resampled to 44 100 Hz)
- At least 2 tracks required

**How it works:**
1. MEGAMI’s diffusion model (FxGenerator) generates audio-effect embeddings from the stems
2. The effect processor (FxProcessor) applies those embeddings to produce a polished stereo mix

Paper: [arxiv.org/abs/2511.08040](https://arxiv.org/abs/2511.08040) | License: CC BY-NC-SA 4.0