A real AI transcription model, running in your browser
Most free transcription tools upload your audio to a server, process it there, and send back the result — meaning your recording, even briefly, passes through a computer you don't control. This tool works differently: it downloads OpenAI's Whisper speech recognition model once to your browser and runs it directly on your own device, so your audio is never transmitted anywhere to be transcribed.
How it works
- Choose an audio file. Drag one in or tap to select it from your device.
- Click "Transcribe." The first time, your browser downloads the speech model — after that, it's cached and starts almost instantly on future uses.
- Download the result. Get a plain text transcript, or an SRT subtitle file with timestamps.
What to expect, honestly
This is genuine machine learning running on your own hardware, not a server farm built for this exact task — so it's worth setting expectations accordingly. The first use downloads roughly 150MB once. Processing speed depends on your device: a short clip on a reasonably modern computer finishes quickly, while a long recording on an older laptop will take real time. Currently, this tool is built and tuned specifically for English speech; accuracy is good for clear audio but will drop with heavy accents, overlapping speakers, or significant background noise, the same as any automated transcription.
Common uses
- Generating subtitles for a video before uploading it somewhere.
- Turning a voice memo or interview recording into searchable, editable text.
- Getting a rough transcript of a meeting or lecture recording.
- Drafting captions for accessibility, then refining them by hand.
If background noise is a problem
A noisy recording can meaningfully hurt transcription accuracy. If your audio has significant background hiss or hum, running it through the site's Audio Noise Remover first can sometimes improve the transcript quality here.
Your audio never leaves your device
Because the model runs locally, this tool never sees or stores the audio you transcribe here. See our Privacy Policy for full details.
Frequently asked questions
+Is this transcriber really free, with no sign up?
Yes. There's no account, no per-minute pricing, and no limit on how many files you can transcribe.
+Do you upload my audio to a server to transcribe it?
No — and this is what makes it different from most transcription services. An AI speech-recognition model (OpenAI's Whisper) is downloaded once to your browser and runs there, meaning your audio is processed entirely on your own device. It's never sent anywhere.
+Why does it take a moment to load the first time?
The first time you use this tool, it downloads the Whisper model — around 150MB. Your browser caches it afterward, so future transcriptions on the same device start almost instantly, without downloading it again.
+Does it support languages other than English?
Not currently — this tool uses an English-specialized version of the Whisper model for the best accuracy and speed on English speech. Multilingual support may come in the future, but for now it's built and tuned specifically for English.
+How accurate is it?
Genuinely good for clear speech with minimal background noise, though not perfect — accuracy drops with heavy accents, overlapping speakers, or a lot of background noise. It's worth reviewing the transcript rather than assuming it's flawless, the same as with any automated transcription tool.
+How long does transcription take?
It depends on your device and the length of the audio — processing happens on your own computer's processor rather than dedicated server hardware, so a longer file or an older device will naturally take longer. A short clip on a modern laptop is typically fast; a long recording on an older device will take a while.
+What do I get — just text, or subtitles too?
Both. A plain .txt transcript, and if timestamps were detected, a properly formatted .srt subtitle file ready to use with most video players and editors.