noScribe — локальная расшифровка интервью с разделением спикеров

★ 2.1K

noScribe is an open-source program for transcribing audio interviews that runs entirely on your own computer. Researchers, journalists and social scientists reach for it when they need an accurate transcript of recorded conversations with speaker separation — who said exactly what. Everything is computed locally: recordings under a confidentiality agreement and personal interviews never go to the cloud. Whisper-based speech recognition and a separate speaker-labelling module sit inside. A built-in editor lets you check the transcript against the audio and correct it. The difference from simple transcription tools is the focus on quality and on separating speakers, which is critical for qualitative research and later analysis. It goes one way: sound to text; recognition takes time, but the result is ready for analysis.