AudioLDM 2
Latent diffusion text-to-audio that turns prompts into sound effects, speech, and music
About
AudioLDM 2 is a latent text-to-audio diffusion model from Haohe Liu's team at Imperial's Centre for Vision, Speech and Signal Processing. It pairs self-supervised audio pretraining (an AudioMAE-style representation) with a GPT-2 language model and a latent diffusion UNet, so a plain text prompt can yield sound effects, human speech, or music. Three official checkpoints ship: the base audioldm2 (1.1B total), audioldm2-large (1.5B), and audioldm2-music trained specifically for music, all available in the diffusers library from v0.21.0. It is a research favorite and a common baseline, but generations are short clips rather than full compositions, and the weights' license terms should be checked on the model card before any commercial use.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: competitive generation quality on standard text-to-audio benchmarks in the paper (arXiv 2308.05734)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: HF model card cvssp/audioldm2 verified: three checkpoints (1.1B / 1.5B / music), diffusers v0.21.0+ availability; License terms not re-verified this session — flagged for model-card check before commercial use; Exact release day not verified — releasedAt null.