VoxVerbatim
Need a hand? We're a small team and we read every message.
The fastest way to reach us is by email at
[email protected]. To help us help
you, please include your Windows version, which speech services you have enabled, and a copy of the
relevant lines from vox-verbatim.log if the app reported an error.
No. It has no account of its own, no analytics, and no telemetry. When you transcribe a recording, the audio is sent only to the speech services you've enabled, using API keys you provide, and never passes through a server of ours. See our Privacy Policy for details.
At minimum, ElevenLabs and OpenAI: every transcript is built from those two, and a run won't start without them. Microsoft (via your own Azure resource) and AssemblyAI are optional extra opinions, and can be switched off on their own Settings pages. Deepgram has a settings page but currently takes no part in a run. Several of these give a new account enough free credit to transcribe a good deal before you've spent anything.
Because nobody sits in the middle. With your own accounts, nobody marks the price up, nobody meters what
you use, nobody keeps a copy of your recordings, and nothing stops working on the day somebody else
stops paying for a server. The cost is that the sign-ups are yours to do; the Settings dialog on
Ctrl+, is where each key goes.
That depends on the services you use and their current prices. Before a run starts, VoxVerbatim shows an estimate worked out from rates you type into the Costs page in Settings yourself, and asks you to confirm before anything is sent, unless you've switched that confirmation off. We don't set or track each service's prices; check their own pricing pages for current rates.
Several services read each recording independently, and where they disagree, VoxVerbatim treats that
disagreement as a signal rather than settling it by guesswork. Amounts, dates, account numbers and
similar values are never decided by scoring: they're left for you in the review window, marked
[UNCERTAIN: ...] in the plain-text export until you settle them.
Yes, as an option you switch on per recording in the Transcribe dialog, not a global setting. Left off, no service is told Afrikaans is possible. Switched on, only ElevenLabs and OpenAI handle it well; Microsoft and Deepgram don't support it, and AssemblyAI's older Afrikaans model is a noticeably weaker second opinion. Switch it on where Afrikaans may genuinely occur in the recording, and leave it off otherwise.
In %LocalAppData%\JB Org\VoxVerbatim, as plain JSON files you can read or edit by hand.
Deleting one costs you what it held and nothing else; a damaged file is ignored rather than stopping the
app from starting.
Not currently. VoxVerbatim is a Windows application.
We'd love to hear it, by email at [email protected], or as an issue on GitHub.