StemFX Models Music Mixing Style as Token Prediction

A paper targeting ISMIR 2026 introduces StemFX, a model that treats music mixing style as an autoregressive token prediction task. Trained on 105,000 songs, the model predicts variable-length FX chains for separated stems.

2026-08-04 ~ 2026-08-04 · 2 related posts

• StemFX frames music mixing as token prediction on 105K songs — affige_yang · 2026-08-04
• ISMIR 2026 paper models music mixing as token prediction with variable-length FX chains — affige_yang · 2026-08-04