SongGen is an open-source music generation model. The user can input lyrics and description information as text, along with an optional audio clip of voice for cloning, with the input audio clip needing to be around 3 seconds. SongGen will output its generated music in an audio file that is up to 30 seconds in length.
Year: 2025
Website: https://liuzh-19.github.io/SongGen/
Input types: Audio Text
Output types: Audio
Output length: 30 seconds
AI Technique: Transformer
Dataset: 8,000 hours of audio from Million Song Dataset (MSD), Free Music Archive (FMA), and MTG-Jamendo Dataset
License type: Apache-2.0 license
Real time:
Free:
Open source:
Checkpoints:
Fine-tune:
Train from scratch:
To run the model locally, follow the steps given here