FAQ
Whimscribe is built on the idea that your audio should stay on your computer. Everything happens inside your browser tab. Nothing is uploaded anywhere.
Privacy & your data
Does my audio leave my computer?
No. Your audio file is read and transcribed entirely inside your browser tab.
Is my audio used to train an AI model?
No. The transcription model runs locally on your machine. It doesn't learn from anything you give it, and nothing you transcribe is sent anywhere for training.
Do you store or log my transcripts?
No. Your transcript only exists in your browser tab and it disappears when you close it unless you copy it or download the .txt file.
Are there any analytics, trackers, or telemetry?
No. This site doesn't include any analytics, trackers or telemetry. Nobody can see what you're transcribing.
How can I be sure of all this?
The whole project is open source. You can read the code on GitHub, or open your browser's developer tools and watch the Network tab while you transcribe — you won't see your audio being sent anywhere.
Internet & offline use
Does it need an internet connection?
On your first visit the app downloads the AI model (about 2 GB). After that, your browser keeps a copy, and you can transcribe with no internet at all.
Where does the model come from?
It currently uses the Cohere Transcribe model, downloaded from Hugging Face (a well known host for open AI models). After it's downloaded, it lives on your computer.
Where is the model stored on my computer?
In your browser's own storage for this site. If you clear your browsing data for this site, the model gets removed and will need to be downloaded again next time.
Browser & hardware
Why does it need a recent browser?
Whimscribe uses WebGPU, a newer browser feature that lets the AI model run on your graphics card at a usable speed. Recent versions of Chrome, Edge, and Safari on a desktop computer all support it.
Why doesn't it work well on my phone?
The model is around 2 GB and needs a fair bit of memory and graphics power to run. Most phones don't have enough of either, so a laptop or desktop is recommended.
Using it
What audio formats can I use?
Most common ones: mp3, wav, m4a, ogg, flac, and webm. If your browser can play it, Whimscribe can likely transcribe it.
Is there a file size or length limit?
There's no hard limit, but very long recordings take longer and use more memory. For best results, try files under an hour or so.
Is it really free?
Yes. There's no account, no payment, no usage limits. The cost of running it (the electricity and the GPU time) is paid by your own computer.
Still wondering about something? The code is open, have a poke around on GitHub or get in touch with me.