Search papers, labs, and topics across Lattice.
2
0
3
12
Achieving state-of-the-art performance in streaming speech-to-speech translation with a mere 2,000 hours of paired data challenges the notion that massive datasets are essential for high-quality results.
Standardized evaluation of nonverbal vocalizations in TTS is now possible with NV-Bench, a new benchmark that treats NVs as communicative acts, not just acoustic artifacts.
