MP3 transcription
Turn an MP3 recording into editable text in your browser. Review timestamped segments, edit the transcript, and export text or subtitles.
Try one MP3 file every rolling 30 days without an account, up to 30 minutes.
MP3
Open a prepared LibriVox audiobook recording to see how an MP3 transcript is presented. Follow its timestamps alongside the audio.
Test the same public-domain recording encoded at 32, 64 and 128 kbps using the same page settings. Compare each transcript with the source audio for missed words and wording changes.
257 KB · WER 0.0% against the original text
No wording differences from the 128 kbps transcript; no visible word errors.
513 KB · WER 0.0% against the original text
No wording differences from the 128 kbps transcript; no visible word errors.
1025 KB · WER 0.0% against the original text
No wording differences; used as the comparison baseline.
Test conditions: LibriVox "A Scandal in Bohemia", one 65.5-second excerpt, mono 44.1 kHz CBR MP3; Workers AI whisper-large-v3-turbo; single run on 2026-11-16, uploaded as MP3. One reader and one excerpt, so differences within noise are not real gaps between bitrates.
Use this table to estimate how large an MP3 will be before you upload it. The table uses specification-based bitrate conversions, not measured upload limits.
| MP3 bitrate | Approximate size per hour | Approximate size for 10 hours |
|---|---|---|
| 128 kbps | 57.6 MB | 576 MB |
| 64 kbps | 28.8 MB | 288 MB |
| 32 kbps | 14.4 MB | 144 MB |
Variable-bitrate files can differ. Paid audio files can be up to 10 hours; free files can be up to 30 minutes.
An MP3 stores audio using compression. Its bitrate, sample rate, channel layout and the sound in the recording describe different parts of the file. They also affect how large an exported MP3 is.
Bitrate describes how much audio data is used over time. A higher bitrate generally creates a larger file at the same duration. Use the size table above to compare steady-rate examples.
Sample rate describes how often the audio signal is sampled. It is usually set when the recording is created or encoded. Keep the source settings in mind when comparing two MP3 copies.
Mono and stereo describe channel layout. Mono suits a single voice; stereo preserves separate left and right channels and their placement in the recording.
Background sound includes room noise, music and other voices. Listen before uploading, then check uncertain words against the audio.
CBR and VBR describe how an encoder allocates data. CBR uses a steady rate; VBR adjusts its rate as the audio changes. A VBR file can differ from the table's constant-rate size estimate.
MP3s can come from recording apps or exports. Play the saved file from beginning to end before choosing it for transcription.
Some apps save in their own format and offer MP3 only when exporting or sharing. If the export settings include channels, mono suits a single speaker; compare the estimated file size before choosing a bitrate.
Use the downloaded episode file when you have permission to process it. Episodes sometimes include music and introductions, so note where the spoken section you need begins.
Export the call as an MP3 from the recording app or service that created it. Confirm that the saved file contains the conversation and that you are allowed to upload and process the recording.
Some voice memos need conversion in their original app or an audio editor. For one voice, consider mono. The size estimate table includes 64 kbps; check the exported audio before relying on that setting.
The transcript is arranged in timestamped segments. SRT and VTT exports pair each text segment with its time in the MP3. Compatible players use those times to display subtitles.
Use SRT or VTT when preparing captions for a podcast video or another player that accepts subtitle files. Podcast editors can use timestamps to organize show notes around sections. For research, interviews or fact-checking, a timestamp marks where to check a quotation against the recording.
Subtitle timing follows the processed audio timeline. If you trim or rearrange the MP3 in another editor, check the exported subtitles against the edited version before publishing.
1 00:00:00,000 --> 00:00:00,340 companion. 2 00:00:01,480 --> 00:00:08,360 One night, it was on the 20th of March, 1888, I was returning from a journey to a patient, 3 00:00:08,400 --> 00:00:13,580 for I had now returned to civil practice, when my way led me through Baker Street.
Check both duration and estimated size before uploading. The table above gives approximate sizes for steady bitrates; variable-bitrate encoding produces different actual sizes.
Paid audio uploads can be up to 10 hours per file. Free files are limited to 30 minutes. If your MP3 exceeds its limit, split it in an audio editor and upload the parts separately.
Note the segment order when splitting a recording so you can put the text back in sequence. If file size differs from the estimate, check its bitrate mode and export settings.
Choose an MP3 in a browser-based transcriber, review the transcript, and export the format you need. This page brings those steps together without a separate audio editor.
Select a file from your device or drop it into the upload area.
The selected MP3 is uploaded for transcription. Keep this page open while the transcript is prepared.
Read timestamped segments, edit wording, and download TXT, DOCX, SRT or VTT.
Copyable workflow
Choose an MP3 → Review timestamps → Edit and export
Unlimited is $10 a month, and hour packs start at $3 for 10 hours, valid for 12 months. See audio transcription plans and hour-pack pricing, including pack validity and billing details.
Transcript segments include timestamps. Export SRT or VTT for subtitle timing, or TXT and DOCX for a text document.
Use an audiobook recording, an MP3 exported from a phone recording, a podcast file you may use, or audio from a call. Confirm you have permission to upload and process it.
At a steady 128 kbps, one hour is approximately 57.6 MB. At 64 kbps, it is 28.8 MB; at 32 kbps, it is 14.4 MB. These specification-based estimates assume steady bitrate and decimal megabytes.
In our test of one 65-second clean audiobook reading, 32, 64 and 128 kbps MP3s produced identical transcripts. Noisy or fast speech may behave differently, so the table above lists the exact test conditions.
Yes. You can transcribe an MP3 with two speakers, but the transcript does not distinguish or label the speakers. Review the recording when you need to determine who said each line.
The upload checks media content, so a mismatched filename extension can work when the audio is readable. If it cannot be read, follow the error guidance.
Uploaded audio is deleted after 48 hours. Guest transcripts stay for one day, free-account transcripts for 30 days, and subscriber transcripts for 90 days.
Use Delete in the result area to remove a transcript sooner.
For other audio files, open Audio to Text.