transcriptionstream

Why we included this project

Teams that regularly need to turn meeting recordings, interviews, or lecture audio into searchable text will find this a handy starting point. Instead of wiring together Whisper and speaker diarization yourself, you get a self-hosted service that runs fully offline and accepts files either through an SSH drop folder or a simple web interface. Output lands in dated folders with time-synced playback and highlighted text for review. It can also hand transcriptions to a local Ollama model for structured summaries that call out speakers, topics, action items, and dates, and it indexes everything with Meilisearch for fast full-text search across past calls or fieldwork. You'll need an NVIDIA GPU and some Docker Compose experience, but for a private transcription pipeline with minimal assembly, it's worth a look.