Questions, problems, and how to reach us.
The more of this you can include, the faster we can help:
Please do not send us a recording or a transcript. We do not need it, and we would rather not receive it — see our Privacy Policy. A description of what went wrong is enough.
No. Transcription, speaker identification, summarisation and search all run on your Mac. Recordings, transcripts and searches are never uploaded, and we have no server that could receive them. The only network connection Verbatim makes is a one-way download of the AI models themselves, from Hugging Face, and that download sends none of your data.
The models are several gigabytes in total, so the first launch takes a while on a slow connection. If a download fails, Verbatim › Settings › Models lets you retry the individual model — you will not have to start the whole set again. Downloads resume rather than restart.
Both are macOS permissions, and macOS remembers your answer:
If you declined a permission earlier, macOS will not ask again — you have to turn it on in System Settings yourself. Verbatim keeps working without either one; it simply cannot record that source.
Inside Verbatim’s own application container on your Mac. Each recording is a self-contained document holding its audio, transcript and summary. Deleting a recording in Verbatim deletes it from disk; removing the app removes everything with it.
Yes. Settings › Models lists every downloaded model with its size, and you can remove any you are not using. The summarisation model is by far the largest, at around 5 GB, and Verbatim transcribes perfectly well without it — you will only be asked for it again if you request a summary.
Purchases are handled by Apple, not by us, so refunds go through reportaproblem.apple.com.
Verbatim is built so that we never receive your content. What we do and do not process is set out in full in our Privacy Policy.