Transcribing multi-channel audio (Channel deorization)#55

Open

slavonnet wants to merge 2 commits intosalute-developers:mainfrom

slavonnet:multichannel-deorization

Commits on Jan 13, 2026

Added functionality for transcribing multi-channel audio with automatic diarization (channel/speaker separation). The method correctly processes overlapping speech and sorts the results by time, en…
slavonnet
committed

Commits on Jan 14, 2026

Fix error - stft(torch.cuda.HalfTensor[1, 270], n_fft=320, hop_length=160, win_length=320, window=torch.cuda.FloatTensor{[320]}, normalized=0, onesided=1, return_complex=1, align_to_window=None) : …
slavonnet
committed