Uploads use TLS in transit and encrypted object storage at rest. Files are automatically deleted after 7 days. We do not use your media to train public models.

Credits remain on your account until used. Processed outputs are retained for 7 days so you can download transcripts, subtitles, and stems — then they are permanently removed.

Transcription uses Whisper-class models, stem separation uses Demucs, denoise uses DeepFilterNet, and optional speaker diarization uses PyAnnote — all running on our on-premise GPU stack for consistent quality without cloud markup.