mp3 → text · free · no install
MP3 to text, in your browser.
Drop the file, read the transcript. Nothing to install, no account, and the audio never touches your clipboard manager.
Getting a clean transcript out of an MP3
The model is the same one OpenAI serves. Almost all of the variance in what you get back comes from the recording, not the software.
- Mono beats stereo. A stereo voice recording is two copies of the same thing at double the file size — collapsing it to mono buys you headroom under the15 MB cap for free.
- Distance is the enemy. A phone on the table across a meeting room loses more accuracy than any codec setting will.
- Crosstalk is unrecoverable here. Whisper transcribes overlapping speakers as one stream. If you need to know who said what, that's speaker labels, which come with Unlimited rather than the free tier.
- Leave the language alone. Detection is automatic and reliable across 99+ languages, including code-switching mid-sentence.
Next steps
Need subtitles rather than a transcript? Export it as .srt → Doing this on more than a handful of files? See what the API costs →
FAQ
How long does an MP3 take to transcribe?+
Most files come back in 5 to 15 seconds. The upload is usually the slow part, not the transcription — a 20 MB file on a weak connection spends most of its time in transit. The result screen breaks out upload time and transcribe time separately so you can see which one you waited on.
Does MP3 bitrate affect accuracy?+
Barely, above about 64 kbps. Speech recognition cares far more about background noise, microphone distance and people talking over each other than it does about compression. A clean 96 kbps voice memo will beat a noisy 320 kbps recording every time.
My MP3 is larger than 15 MB. What now?+
Try it anyway — the cap is checked after your browser downsamples the audio, so a bigger MP3 often still fits. If it doesn't, three options. Split it — most editors and ffmpeg will cut an MP3 without re-encoding. Re-export it as mono at a lower bitrate, which usually cuts the size by half or more with no accuracy cost. Or get Unlimited ($9 a month), which takes files up to 50 MB and 2 hours; if you're writing code, the API's practical ceiling is how long you're willing to wait rather than a hard cap.
Can I transcribe an MP3 podcast episode?+
Yes, that's one of the most common uses here — along with interviews, lectures, voice memos and recorded calls. A full episode usually runs past the free 30-minute limit, so either split it, use Unlimited (files up to 2 hours), or run it through the API.
What about formats other than MP3?+
wav, flac, m4a, aac, opus and ogg all work, and so do video containers like mp4, mov and webm — the audio track is extracted for you. If ffmpeg can decode it, so can this.
Can I batch a folder of MP3s?+
Not in the browser — it's one file at a time, 3 a day on the free tier. Batching is what the API is for: a loop over a directory with the OpenAI SDK is about ten lines, and there's a working example on the Python page.