Turns folders of recordings into transcripts you can trust.
An accessibility-first Windows app, built for JAWS, NVDA and ZoomText. Point it at a folder of recordings and it asks several speech-to-text services the same question at once, combining their answers into one transcript and flagging the words they couldn't agree on, rather than guessing.
Choose a folder of recordings and VoxVerbatim lists every file with its length and size. Listen to any of them, write louder copies of the ones too quiet to hear or transcribe cleanly, and transcribe the ones you check. For each recording, several speech services read it at once and their answers are laid side by side: the spelling, the timing and the speaker can each come from a different service, whichever answered that question best. Where the evidence doesn't settle something, such as an amount or a name, the transcript says so and waits for you, instead of quietly picking a guess.
VoxVerbatim is free to run and holds no account of its own with any speech service. You sign up with the services yourself, paste your own API keys into Settings, and each service bills you directly for what it does. There is nothing in the middle marking the price up or keeping a copy of your recordings.
VoxVerbatim is free and open source under the MIT License. Everything you need for transparency and assistance is here.