Project ◆ Hidden gem
AudioLDM2
haoheliu/AudioLDM2 · Python · MIT
Latent diffusion for text-to-audio and music
AudioLDM 2 remains a solid open text-to-audio baseline for SFX and short musical phrases.
8.3 / 100 · Score
★ 2.6k Stars
+2 30-day growth
25 min Time to first run
Not required API key
Score
Trend 26
Activity 8
Quality 65
Community 19
Relevance 100
Setup at a glance
DeployCli, Web
GPU8 GB
API keyNot required
UI languagesen
Doc languagesen
Install
pip install git+https://github.com/haoheliu/AudioLDM2.git Deep dive
AudioLDM 2 is a text-to-audio generation system built on latent diffusion: describe a sound, a music snippet or a speech effect and it synthesizes the audio — no recordings required.
The 2023 update added two new checkpoints: a 48kHz model for high-fidelity output and an improved 16kHz model trained on more data, so you can trade quality against speed per use case.
It integrates with the diffusers pipeline API (with a Colab walkthrough), which makes it one of the easiest research audio models to embed in an application.
Use cases
Game SFX from prompts
Short music beds
Audio research baselines
Good for / Not ideal for
Good for
Game audio prototyping
Not ideal for
Full 3-minute vocal songs