Adam Rotmil

I fine tune speech models on small, noisy, irreplaceable audio archives. Doing detailed manual audio restoration to make magic happen in the training data! Current interests: LoRA on autoregressive TTS, diarization and ECAPA speaker verification at corpus scale, register/era conditioning, reference anchored long-form synthesis, style transfer adapters, and self distillation on filtered outputs. Speech models in English, french, german, japanese.