AudioLDM 2
Generate sound effects and audio scenes; a separate checkpoint targets music.
GenerationOpen-source code
- Start here
- Try a clearly described short audio scene. Compare several seeds before spending time tuning a vague prompt.
- Training data
- The card reports 1.15 million training hours for general audio and 665,000 for the music variant. Consult the paper for component-specific corpora.
- Training rights
- The card is not a recording-level rights inventory. Pretrained encoders and source corpora carry separate provenance.
- License & output terms
- Linked base weights: CC BY-NC-SA 4.0.
- Limitations
- Non-commercial, share-alike weights. Prompt and seed sensitivity; short audio generation does not ensure coherent song structure.
- Technique
- Self-supervised features + latent diffusion
- Runtime
- Local diffusion inference through the upstream implementation or Diffusers.
- Parameters
- Not itemized for this release/family. Parameter sizes shown elsewhere may exclude encoders, decoders or other runtime components.
Sources reviewed . Training and rights notes summarize provider disclosures. A model license does not clear every source recording or generated output. Parameter counts refer to the named variant or component, not total runtime memory.