What audio formats work?

MP3, WAV, M4A, OGG, and FLAC all work, and generally anything your browser can decode. AMR files from phone recorders also work: browsers have no decoder for them, so the tool brings its own and runs it locally. Older Windows WMA files remain unreliable; convert those to MP3 or WAV first.

Is there a file size or length limit?

There is no hard size limit. Long files are transcribed in parts automatically, but the whole recording is decoded into memory before that starts, so about an hour per file is the comfortable ceiling. Longer recordings can be split into parts named in order and selected together, and they are joined into one transcript.

It uses the Whisper speech-recognition model. Accuracy is strong on clear audio and improves with the larger model option, at the cost of a bigger first download. You can also edit the text in the page before downloading it.

Does a higher bitrate MP3 transcribe better?

Up to a point. Very low bitrate speech recordings lose detail the model needs, so the transcript degrades. Above roughly 128 kbps for speech there is little further gain, because the model works from a 16 kHz mono version of your audio regardless. Re-encoding a low quality file to a higher bitrate recovers nothing.

Can I transcribe a WhatsApp voice note?

Yes. Exported WhatsApp voice notes are usually OGG Opus files, which browsers decode natively, so you can drop one straight in.

Your audio never leaves your device

Convert MP3 to text, free and private

Turn an MP3 into a clean text transcript in your browser. No signup, nothing uploaded, and it also handles WAV, M4A, OGG, and FLAC.

Drop audio here, or click to choose

MP3, WAV, M4A, OGG, FLAC, AMR. Everything runs on your device. Your audio is never uploaded.

Got a long recording in several files? Select them all at once. Name them in order, like meeting_part1.mp3 and meeting_part2.mp3, and they will be joined into one transcript.

QualityLanguageTime-stampLabel who is speaking How many people talking? Sharper speaker labels Clean up the audio first

What is this recording about?

Takes a little extra time, and your audio still never leaves your device. Labels go by how voices sound, so treat them as a best effort: one person can be split across two or three labels, similar voices can be merged, and someone joining partway through may be folded into a speaker already talking. Giving a number usually helps. Check them before relying on them.

Specifying this number improves accuracy. Count people who really take part, whenever they join. Someone who only says a word or two is better left out.

Timestamps each word instead of each phrase, so short interjections like "Yeah" go to the right person. On a busy four-person recording this corrected about 1 word in 5; it changes little when people talk in long turns. Adds roughly 40% to the time on a computer, and almost nothing on a phone.

Phone audio loses the high frequencies that separate consonants, so they are put back before transcribing. Only phone-quality recordings are treated, and you are told when it happens. Also even out volume is for one case: someone far from the microphone and much quieter throughout. It cost accuracy on other recordings, so use it only if a quiet person came out badly.

Worth filling in. The speech model knows ordinary English, not your subject, so a word it has never heard becomes whatever sounds closest, and it usually gets that word wrong every single time it comes up. A sentence is enough, and the names and jargon in it are picked out and watched for. It cannot guess a word you do not mention, so name the people and the terms that matter.

Free, no signup. The first time you use it, it takes a minute or two to set up. After that setup is quick. Works with MP3, WAV, M4A, OGG, FLAC and AMR, and long recordings are handled in parts.

MP3 to text without uploading the file

Most MP3-to-text converters ask you to upload your file to their servers. MeetingScribe does the conversion in your browser instead, so your recording stays on your device. That keeps private audio private, and it also means there is no account, no upload queue, and no size cap imposed by somebody's free tier.

How to convert

Drop your MP3 into the tool above, or click to choose it.
Pick a model size and language, or leave them on the defaults.
Optionally turn on speaker separation to label who is speaking.
Click Transcribe, then edit the text and download it as a .txt file.

Which formats work, and why some do not

This tool relies on your browser's own audio decoder, which is why the supported list looks the way it does. If a browser can play it, this can transcribe it.

MP3. Universal, works everywhere, the safest choice for sharing a file between devices.
WAV. Uncompressed, so the largest files and the best quality. Ideal for short recordings where accuracy matters most.
M4A. What iPhone Voice Memos, Zoom's audio-only export, and most modern recorders produce. Works directly, no conversion needed.
OGG and Opus. Common for WhatsApp voice notes and web recordings. Decodes natively.
FLAC. Lossless compression, so full quality at roughly half the size of WAV.

The ones that need converting first:

3GP. Produced by some older Android call recorders. Browsers cannot decode it, so convert to MP3 or WAV with a desktop converter first. Plain AMR files, the other format those recorders write, now work directly: the tool carries its own AMR decoder.
WMA. An older Windows format with patchy browser support. Convert it.
Video files such as MP4 and MOV. Not audio, so extract or export the audio track first, then transcribe that.

Converting without hunting for a converter

Most online converters ask you to upload the file, which is the thing you came here to avoid. If you would rather not, ffmpeg is a free command-line tool that runs on your own machine, and three commands cover every case above. Replace the input name with your own file.

Pull the audio out of a video, without re-encoding it, so nothing is lost and it is quick:

ffmpeg -i meeting.mp4 -vn -acodec copy meeting.m4a

Turn an old WMA or 3GP recording into a WAV the browser can read. The extra flags ask for 16,000 samples per second in one channel, which is exactly what the speech model works at, so the file is as small as it can be without losing anything the model would have used:

ffmpeg -i interview.wma -ar 16000 -ac 1 interview.wav

Split a recording too long to hold in memory into half-hour parts, named in order so they can be selected together and joined back into one transcript:

ffmpeg -i long.mp3 -f segment -segment_time 1800 -c copy part%02d.mp3

WAV is worth knowing about for a second reason. It is the one format read in slices rather than decoded whole, so the memory ceiling that applies to long compressed files does not apply to it. A recording that fails as a three-hour MP3 will usually go through as a WAV.

Common sources, and what to expect from each

Phone voice memos. Usually M4A, and usually good quality if the phone was near the speaker. This is the easiest case and the base model handles it well.
Recorded phone calls. Narrow-band audio, often heavily compressed, sometimes with both sides at different volumes. Use the small model, and expect speaker separation to struggle if both voices arrive on one channel.
Dictation. The best case for this tool. One clear voice, close microphone. Turn speaker separation off for a faster run and a cleaner transcript.
Lecture or conference recordings from the back of a room. Reverberant and distant, which is genuinely hard. The small model helps, but no tool recovers detail the microphone never captured.
Podcasts and interviews recorded properly. Excellent results, and a good candidate for subtitle export as well.

Choosing a model for your file

The model choice matters more than any other setting, and the right answer depends on the audio rather than on the length:

tiny, around 50 MB. For clean single-voice recordings where you want a fast rough draft you will edit anyway.
base, around 90 MB. The default, and correct for most well-recorded speech.
small, around 270 MB. Worth the download for accents, background noise, phone audio, distant microphones, or anything you intend to quote from.

The download happens once and is then cached, so the cost of choosing small is a slower first run rather than a slower every run. If you are unsure, transcribe two minutes of the file with base, look at the result, and decide from there.

Getting more out of the text

The same transcription also produces subtitle files. Open the Subtitles tab to download SRT or VTT with timestamps, which is what you want if the MP3 came from a video. See audio to SRT and VTT for the format differences and how to load them.

For a summary rather than a transcript, the Copy notes prompt button puts your transcript plus clear instructions on your clipboard so you can paste it into any AI chat. Only the text goes, and only if you choose to paste it.

Last reviewed: August 2026

Frequently asked questions

What audio formats work?: MP3, WAV, M4A, OGG, and FLAC all work, and generally anything your browser can decode. AMR files from phone recorders also work: browsers have no decoder for them, so the tool brings its own and runs it locally. Older Windows WMA files remain unreliable; convert those to MP3 or WAV first.
Is there a file size or length limit?: There is no hard size limit. Long files are transcribed in parts automatically, but the whole recording is decoded into memory before that starts, so about an hour per file is the comfortable ceiling. Longer recordings can be split into parts named in order and selected together, and they are joined into one transcript.
How accurate is it?: It uses the Whisper speech-recognition model. Accuracy is strong on clear audio and improves with the larger model option, at the cost of a bigger first download. You can also edit the text in the page before downloading it.
Does a higher bitrate MP3 transcribe better?: Up to a point. Very low bitrate speech recordings lose detail the model needs, so the transcript degrades. Above roughly 128 kbps for speech there is little further gain, because the model works from a 16 kHz mono version of your audio regardless. Re-encoding a low quality file to a higher bitrate recovers nothing.
Can I transcribe a WhatsApp voice note?: Yes. Exported WhatsApp voice notes are usually OGG Opus files, which browsers decode natively, so you can drop one straight in.

Audio to subtitles Speaker labels How it works, technically Private transcription