After its one-time setup, CaptionLab's transcription step makes no network requests at all — it runs entirely against a local AI model already installed on the machine.
This page describes CaptionLab's technical network behavior honestly, not a formal certification. CaptionLab is not independently certified or tested for classified/air-gapped government environments — if you work in one, verify against your own environment's specific security requirements before relying on it.
Once installed, generating subtitles doesn't attempt any outbound connection.
The Whisper model, Python runtime, and ffmpeg all run from local files installed during setup.
The one thing that does reach out to the internet is an occasional license validity check — separate from transcription itself, and can go a couple weeks offline before it's required.
Installing CaptionLab's AI engine downloads roughly 3GB (Python runtime, ffmpeg, the Whisper model) from the internet the first time. That step needs connectivity. Everything after that — every transcription run — does not.
One-time $39. Works offline. 14-day money-back guarantee.
Get CaptionLab →