3 ms·
Ultrafast AI music generation (with latent consistency model)
- tonyabracadabra 2y agoAudioLCM needs only 2 iterations to synthesize high-fidelity audios, while it maintains sample quality competitive with state-of-the-art models using hundreds of steps! The methodological unification of AI studies across different modalities is unavoidable.