Extract the main audio track from an MP4 and save it as a widely supported 192 kbps MP3. Free with clearly published fair-use limits — no sign-up, no watermarks.
Drop your MP4 file here
or click to browse — One file per media job — 100 MB / 10 minute limit
☁️ Server conversion — your file is uploaded securely (HTTPS), converted, and removed from temporary processing storage after the request finishes. Forced-crash leftovers are removed by recovery sweeps. A verified codec/duration report is included on the download page. Which tools upload, and which do not
Lectures, interviews, meetings and music clips are often carried inside video even when the picture is no longer useful. Removing the video creates a much smaller audio-only file for listening, transcription or sharing. The first audio stream is decoded and encoded to 192 kbps stereo MP3; a video with no audio is rejected instead of producing silence.
This page has a real two-second fixture made from generated motion, colour and tones — no copyrighted clip and no visitor data. Download the source, run it through the tool, and compare it with the expected profile-1.0.0 output.
⬇ Sample input (MP4) ⬇ Verified expected output
| Check | Input | Expected output |
|---|---|---|
| Codecs | h264 + aac | mp3 |
| Duration | 2 s | 2.005 s (Δ 0.005 s) |
| File size | 89,106 bytes | 49,581 bytes |
| SHA-256 | 076cee70b78122b5d9197f57a1eb5fa1cd2873e167ce437bb0af831b6697b973 | 0b400f5efcaefcb79da7bfe5d284a345b5aa238b84d81e64d297645bd86acf27 |
| Measured time | — | 87 ms in the published environment |
| Validation | Signature + ffprobe | Codec, complete decode and duration: pass |
No confidence score is invented: transcoding is deliberately lossy, so the defensible result is a set of pass/fail structural checks, not “97% accurate.” Your download page exposes the same per-job evidence report as JSON and CSV. See the method, limitations and six-case comparison, or download the anonymous CSV / JSON dataset.
“Accuracy” for this lossy task means the published structural gates below; perceived quality still needs human review of the output.
| Profile | Date | Validation change | Re-test | Result |
|---|---|---|---|---|
| 1.0.0 | 2026-08-16 | Fixed codec profile; EBML DocType / ISO brand checks; rotation-aware dimensions; ffprobe; complete decode; duration tolerance | FH-MEDIA-03 plus the six-profile release gate and adversarial regressions | pass |
MP4 is the safest delivery container for video. The widely supported combination is H.264 picture with AAC sound, which plays on current phones, browsers, televisions and editing apps. MP4 can contain other codecs too, so a genuine compatibility conversion must inspect and encode the streams rather than merely rename the extension.
MP3 is the most broadly compatible compressed audio format. Music players, car stereos, browsers, editors and upload services understand it. It is lossy: converting audio to MP3 removes some information to reduce size, so the original recording remains the right master for future editing.
| MP4 | MP3 | |
|---|---|---|
| Type | Video/audio container | Audio |
| Compression | Usually H.264/H.265 + AAC | Lossy |
| Transparency | Usually no | — |
| Animation | Video | — |
| Typical use | Sharing, uploads, playback, editing | Listening, transcription, sharing |
A one-hour lecture, interview or meeting may occupy hundreds of megabytes as video even when it will be consumed with the screen off. Extracting its audio creates a portable file for a music player, transcription service or low-bandwidth share. Only the first audio stream is used, and the video contributes nothing to the MP3.
The operation is extraction plus encoding, not restoration. If the MP4 contains thin, noisy or heavily compressed sound, a higher MP3 bitrate cannot invent clarity; it can only avoid adding more damage than necessary. Listen at the beginning, middle and end, verify speech follows the expected timeline, and retain the MP4 whenever the visual record still matters.
Yes — there is no sign-up, payment or watermark. The published safeguards are 100 MB input, a 160 MB output cap, a 10 minute selected-stream limit, an 8 minute request-wide deadline and four jobs per visitor per 10 minutes; they are technical limits, not a paid tier.
Media files are handled one at a time. That makes the documented CPU, memory and cleanup limits enforceable and gives the next file a clean job directory. Finish or clear the current file before adding another.
Yes. Native FFmpeg is too large and resource-intensive for a reliable browser-only conversion, so this tool sends the file over HTTPS to our server. It is limited to 100 MB and 10 minutes and uses an isolated temporary job directory. Normal cleanup removes that directory after success, failure, cancellation or timeout; startup and five-minute sweeps remove a forced-crash leftover once it is older than the request deadline. Media files are never intentionally retained.
No. MP3 is audio-only. The first audio track is decoded and re-encoded; the video, cover art, chapters and additional streams are intentionally discarded.
filehugger.com