Assistance

Report a problem

Open an issue in the repository with the file format, its length, the browser version, and whether the run was local or Pro. A failing job is far easier to fix with those four facts than without them.

Common answers

• The first run is slow: the speech model downloads once, then runs offline.
• A long file takes a while: it is split into windows and transcribed in order, so time scales with duration.
• Nothing is uploaded on a local run — if you need the faster pass, choose it per file.