Files
Rohit Ghumare 6937debf64 feat(phase-08/11): audio generation
Bigram codec-token model with per-style conditional training, temperature sampling, VALL-E-style 3-second prompt continuation. Covers Encodec / DAC / SoundStream + token-AR vs flow-matching production stacks.
2026-04-23 00:16:40 +01:00
..
2026-04-23 00:16:40 +01:00