---
title: SongFormer
emoji: 🧩
colorFrom: yellow
colorTo: green
sdk: gradio
sdk_version: 6.17.3
python_version: 3.10.13
app_file: app.py
pinned: false
license: cc-by-nc-4.0
short_description: "Music structure: section boundaries and labels"
tags:
  - category:analysis
  - subcategory:music-analysis
  - input:audio
  - output:audio
  - output:labels
  - sample-rate:24000
  - channels:mono
preload_from_hub:
  - ASLP-lab/SongFormer SongFormer.safetensors a75880ed1b7375ac71860ec6c4fc9c899cf99515
  - minzwon/MusicFM msd_stats.json,pretrained_msd.pt 4513b38bc25ad1d227b1980819b9691ba97f4d87
  - OpenMuQ/MuQ-large-msd-iter config.json,model.safetensors 0562a57814f6f8bbd9fdea0a25921a2fce1a841a
  - facebook/wav2vec2-conformer-rope-large-960h-ft config.json
---

# SongFormer

SongFormer identifies sections in a music recording using MuQ and MusicFM audio features.
Upload a recording of at least five seconds. Each section comes back as a label over the
recording, running from its start to the next boundary.

The MuQ weights are licensed under CC BY-NC 4.0. Original code and checkpoint links are in
[SOURCES.md](SOURCES.md).

Gradio is held at 6.17.3, the newest release whose `huggingface-hub` floor admits the transformers 4.53.0 the model needs.