Transcribe audio, diarize speakers, and review via web UI
This tool turns any audio file or YouTube link into a speaker‑labeled transcript. It runs Whisper locally (or via OpenAI) to get word timestamps, then segments sentences, embeds each clip with TitaNet, and clusters speakers using UMAP + KMeans. A lightweight web UI lets users inspect, listen, and correct speaker labels. It’s ideal for interviewers, meeting organizers, or researchers who need accurate, editable transcripts.
View on GitHub →Novia-RDI-Seafaring/transcriber