---
title: MelodyFlow
emoji: ✏️
colorFrom: pink
colorTo: purple
sdk: gradio
sdk_version: 6.17.3
python_version: "3.10"
app_file: app.py
pinned: false
license: cc-by-nc-4.0
short_description: Text-guided music editing with flow matching
preload_from_hub:
  - facebook/melodyflow-t24-30secs
  - t5-base
tags:
  - category:generation
  - subcategory:editing
  - domain:music
  - input:text
  - input:audio
  - output:audio
  - sample-rate:48000
  - channels:stereo
---

Text-guided music editing with flow matching, using
[MelodyFlow](https://github.com/facebookresearch/audiocraft), accessible in HARP.

`audiocraft` is vendored from the authors' Space,
[facebook/MelodyFlow](https://huggingface.co/spaces/facebook/MelodyFlow), whose fork moves it to the
torch release ZeroGPU requires. The code is MIT (`LICENSE`) and the weights are CC-BY-NC-4.0
(`LICENSE_weights`), so results cannot be used commercially.

Gradio is held at 6.17.3, the newest release whose `huggingface-hub` floor admits the transformers 4.x audiocraft needs.