Privacy Policy
DOTE Whisper runs entirely on your computer. It decodes, transcribes and labels your recordings locally and never uploads them.
What is stored
Nothing about you. Everything the app keeps is on your own computer.
| Data | Where it lives |
|---|---|
| Transcripts (JSON, SRT and WebVTT files) | Beside the original recording when Auto-save next to source is on, otherwise wherever you save them. |
| Preferences: model, speaker settings, tuning values and window position | In the app's settings folder on your computer. |
| Downloaded models | In the models folder shown in Settings. By default that is inside the app's settings folder. |
| A diagnostic log of errors, warnings and update checks | In the app's log folder. It does not contain transcript text. |
| Temporary working copies of the audio | In your system's temporary folder while a transcription runs. They are deleted afterwards, and any left by a crash are cleared the next time the app starts. |
Your recordings
The app decodes your recordings with a copy of ffmpeg that ships inside it, then runs speech recognition (whisper.cpp) and speaker separation (sherpa-onnx) on your computer. It only ever reads your original files and never changes them. Nothing about their contents is sent to us or to anyone else.
When the app goes online
| When | What happens |
|---|---|
| When the app starts, and when you choose File → Check for updates… | The app asks the DOTE update server at Aalborg University (license.dote.aau.dk) for the latest version number and its release notes. If a newer version exists, the app tells you. It never downloads or installs anything by itself. Without an internet connection the check fails quietly. |
| The first time you use a speech-recognition model | The model downloads from HuggingFace (huggingface.co). The app shows its size and asks first. |
| The first time you turn on speaker labels, or pick a different speaker model | The speaker models download from HuggingFace and from the sherpa-onnx project's releases on GitHub. |
| When you add a custom model | The app asks HuggingFace which files the model you named contains, then downloads them. Browse on HuggingFace opens the HuggingFace website in your web browser. |
These requests carry what any web request does, such as your IP address. The privacy policies of Aalborg University, HuggingFace and GitHub apply to them. DOTE Whisper does not sell data, show ads, or share information with advertisers.
This website
This site has no analytics or cookies. It loads fonts from Google Fonts, and the download page asks GitHub which installer files are available so its buttons can link straight to them. Your light or dark theme choice is saved in your own browser and nowhere else.
Terms & Conditions
By downloading and using DOTE Whisper, you agree to these terms. We've tried to keep them fair and readable, so please do read them.
License to use
DOTE Whisper is free under the MIT License, for personal, academic, commercial and institutional use. The open-source parts it is built on keep their own licenses: whisper.cpp (MIT), sherpa-onnx (Apache 2.0) and ffmpeg (LGPL/GPL). Each model you download also keeps its own license. The standard Whisper models are MIT-licensed; community models from HuggingFace vary, so check the model's page before you rely on it.
Your responsibilities
- Check your transcripts. Automatic transcripts and speaker labels contain mistakes. You are responsible for reviewing them before you quote, publish or otherwise rely on them.
- Keep your originals. The app never modifies source recordings. Keep them, since the recording, not the transcript, is the primary record.
- Use recordings you're entitled to. You're responsible for having the right to process the recordings you use, and for following any consent, ethics or data-handling rules that apply to them and to the transcripts you make.
- Use the software lawfully. Don't use DOTE Whisper for anything unlawful, including recording or transcribing people without the consent the law requires.
No warranty
The software is provided "as is", without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and non-infringement. We do not guarantee that the software will be uninterrupted or error-free, or that any transcript or speaker label will be correct.
Limitation of liability
To the fullest extent permitted by law, the authors and contributors of DOTE Whisper shall not be liable for any data loss or for any direct, indirect, incidental, special, or consequential damages arising out of the use of, or inability to use, the software, even if advised of the possibility of such damages.
Contact
Questions about privacy or these terms? Open an issue on the project's GitHub repository.
Your recordings stay on your computer
Download DOTE Whisper and start your first transcript.
Download free