Speech to text.
Transcribe a recording, voice note, meeting or video into text, or record straight from your microphone. The transcribing happens in your browser, so your audio stays with you.
Add a recording, press Transcribe, get the text
The transcriber runs OpenAI’s open-source Whisper speech model inside your browser, so the recording is processed on your own device.
Add a recording
Upload audio or video, or record from your microphone.
Transcribe
Choose the language (or let it detect it) and press Transcribe.
Download
Save the transcript as text, or as .srt subtitles for a video.
Why your recording never leaves your device
Most transcription services upload your audio to their servers. Altto runs the speech model inside your browser instead, so interviews, meetings and voice notes stay private.
-
Nothing is uploadedYour audio is decoded and transcribed on your own device.
-
Subtitles includedDownload an .srt file with timings, ready for video editors and YouTube.
-
No account, no minutes limitTranscribe as much as you like, free.
recordings ever uploaded to our servers. The speech AI runs on your device.
Will it run on my device?
Because the AI runs on your own device instead of on a server, your computer does the work — that’s what keeps your files private. Most computers from the last few years are fine, but a more powerful one is noticeably faster.
Minimum
- ComputerAny laptop or desktop from about the last 6 years
- Memory (RAM)4 GB
- BrowserThe latest Chrome, Edge, Firefox or Safari
- QualityUse “Fast” (about 40 MB download)
Without graphics acceleration, a recording can take as long to transcribe as it does to play.
Recommended Fastest
- ComputerA recent laptop or desktop (e.g. Intel Core i5 / AMD Ryzen 5 or better)
- Memory (RAM)8 GB or more
- BrowserChrome or Edge, which can use your graphics chip (WebGPU)
- Quality“More accurate” (about 80–150 MB download)
Phones can transcribe short voice notes; for long recordings use a computer.
Tested on a mid-range PC with AMD Ryzen built-in graphics (no separate graphics card), in Chrome: An 11-second clip transcribed in about 3 seconds once the model was downloaded (roughly 20 minutes for an hour-long recording on the same PC).