Search papers, labs, and topics across Lattice.
Tencent Ethereal Audio Lab
1
0
3
Forget flat, lifeless speech: this model uses self-critique to generate expressive speech rivaling GPT-4o-Audio, even with significantly less training data.