Why Transcribe exists
Speech recognition got good enough to run on an ordinary PC. That made a simple promise possible: your recordings never have to leave it.
The idea
Most transcription services work the same way: you upload a recording, a server turns it into text, and you trust the service with what was said. For a lot of recordings that is fine. For a client meeting, an interview with a source, a research participant or a doctor's dictation, it is a real question who else ends up with the audio.
Open speech recognition models such as OpenAI's Whisper changed what a normal PC can do. Transcribe packages them into an app that is pleasant to use every day: one library for your recordings, an editor tied to the audio, meeting recording and clean exports, with nothing to upload and no account to create.
Principles
Your data stays yours
Recordings, transcripts, notes and settings stay on your PC. No account, no telemetry, no analytics, no ads.
Works offline, really
The speech models are installed with the app. Transcription never needs the internet, and the app downloads nothing while it runs.
Honest about limits
Prices and renewal terms are stated plainly. Accuracy claims come with the method and the data, and the app says when you should check a passage.
Made for long reading
A calm interface built around the transcript, the audio and the keyboard, in light and dark.
Who makes it
Transcribe - Offline is made by nawki, an independent developer in Türkiye. There is no investor to answer to and no data business behind the app: it is paid for by the people who choose Pro. The legal details are in the legal notice.
Technology and credits
Transcribe stands on excellent open work, credited in full in the third-party notices that ship with the app.
| Part | What it does | Licence |
|---|---|---|
| Whisper, by OpenAI | The speech recognition models: small for Standard, large-v3-turbo for Accurate | MIT |
| whisper.cpp | Runs Whisper efficiently on the processor or, through Vulkan, the graphics chip | MIT |
| WeSpeaker ResNet34 | The speaker model behind speaker detection | CC BY 4.0 |
| Geist and Geist Mono | The typefaces of the app and this website | SIL Open Font License 1.1 |
| Lucide | The icons of the app and this website | ISC |
Contact
Questions, ideas and problems: yusufnawki@gmail.com, or Settings › About › Send feedback in the app. Help and support answers the common questions.
Your recordings stay on your PC.
Free to download. Transcribes without an account, without uploads and without the internet.
Coming soon to the Microsoft Store · Windows 10 and 11