PrivateMP3

Screen recording to MP3

A screen recording is a video file. This guide covers where each recorder saves it, what to check before converting, and what the converter does with it.

To extract the sound, open the recording in the converter. This guide covers where each recorder saves the file, what to check before converting, and what the converter does with a recording that holds more than one audio track.

A screen recording is not an MP3

The capture is a video container — MP4, MOV, MKV or WebM — with a picture track and at least one audio track. Extracting audio writes a new sound file. It does not convert the recording itself into an MP3, and it does not fetch a file from a website. Open a file already on this device in the converter, then press Convert and Save file.

Find the recording on your device first

The recorder decides the container, and the container decides which export paths are open. Look at the file before opening it here.

After the file opens, the File check panel under the waveform shows what the recorder actually wrote: the container, every audio track with its name, codec, channels, sample rate and measured bitrate, which export path the current choice will take, and the expected size of each output. If the recording has no audio track, the page says so before any conversion starts, which usually means audio capture was off when recording. The test results further down list what that panel reported for six test files and how long each export took.

What a two-track recording looks like: the sample

The converter has an Open the sample recording button under the drop zone, for when there is no recording at hand. The sample is a generated test file, not the output of a real recorder: 20 seconds of Matroska, 474,107 bytes, made with FFmpeg from a 320×180 test pattern and two audio tracks. Track 1, named System audio, is AAC stereo encoded at 96 kbps and carries a 220 Hz tone with quiet pink noise. Track 2, named Microphone, is Opus mono encoded with a 32 kbps target and carries a 440 Hz tone that sounds for 1.8 seconds out of every 3, standing in for speech with pauses. Its SHA-256 is 8880b2c704b8aa42410cbc22e135b36ebe845c1ef3fce03172b2aa54503f0c41.

Opened in desktop Chrome on 13 October 2026, File check listed both tracks by name with language, codec, channels and the 48 kHz sample rate, and measured 96 kbps for System audio and 46 kbps for Microphone. The second figure is above the 32 kbps the file was encoded with, because File check adds up the packets that are actually in the file rather than reading a setting. These are the files the converter saved:

TrackOutputSaved size
1. System audio (AAC stereo)MP3, 192 kbps481,704 bytes
1. System audio (AAC stereo)M4A, unchanged copy250,276 bytes
1. System audio (AAC stereo)WAV, 16-bit3,844,076 bytes
2. Microphone (Opus mono)MP3, 192 kbps481,128 bytes
2. Microphone (Opus mono)M4A, new AAC encode485,294 bytes
2. Microphone (Opus mono)WAV, 16-bit1,921,340 bytes

Three things in that table apply to real recordings too. The MP3 is the same size for both tracks, because the converter has one MP3 rate and a mono source does not lower it. The unchanged M4A copy is offered for the AAC track and not for the Opus track, which has to be encoded again, so its M4A is about twice the size of the copy. WAV halves for the mono track, because WAV size is set by channels, sample rate and duration, not by how the source was compressed.

The input file is not copied into memory

A long screen capture is handled differently from a conventional whole-file converter. PrivateMP3 asks the local file for ranges as they are needed. It does not decode the picture frames. For the waveform, decoded audio is reduced to minimum and maximum peaks for time buckets rather than retained as full PCM. This makes the page useful for a lecture you recorded, a meeting captured with permission, a tutorial you made or another local screen recording where only the sound is needed.

The output can be the memory problem

The completed audio still has to be stored somewhere. PrivateMP3 normally streams it into the Origin Private File System, browser-managed storage on the same device, and exposes it through Save file when finished. If that interface is unavailable, the worker falls back to an in-memory output buffer. A long result is more demanding on that fallback even though the source continues to be read in ranges.

Duration still costs time and storage

The browser must decode the audio timeline, build waveform peaks and encode every kept sample. Device speed, free storage, audio codec and output format therefore matter. The picture resolution is not sent to the audio encoder, but the container still has to be read; two recordings of equal duration should not be promised equal processing times.

Keep the tab open and the device awake. The page requests a screen wake lock while converting when the browser allows it, but system power controls can still suspend a background tab.

Check the tracks before converting

A capture may store microphone and system audio as separate tracks, or mix them into one. When several audio tracks exist, a picker appears with codec and channel information. PrivateMP3 can select one track. It cannot combine tracks, separate speakers from a mix or recover audio that the recorder never captured. An interrupted recording may also be missing packets or final container metadata.

Choose the output for the job

For a recording, the File check panel shows the size of MP3, M4A and WAV before Convert is pressed, and says whether M4A will be an unchanged copy. What each output keeps is set out in the size and output guide.

Remove an empty introduction, long break or unused ending before export. The output-duration estimate updates to reflect what remains. To check whether the result will fit a size limit before converting, use the size calculator. Removing time does not always make the file smaller; the guide to smaller lecture audio has the rule and the measurements.

Measured: six generated recordings through the converter

How long does it take to get the sound out of an hour of screen recording, and how big is the result? These are measured numbers from the shipped tool, not estimates.

What was tested

Six files were made with FFmpeg 9.0.1 from its built-in test sources: a moving 30 fps test pattern for the picture and a 220 Hz tone mixed with quiet pink noise for the sound. They are not captures from any particular recorder. Each one copies a container and track layout that screen recordings commonly use, so the results describe how this tool handles those layouts, not how a given recording app behaves.

FilePictureAudio as writtenSize
A. 60 min MP4H.264, 1920x1080AAC, stereo, 48 kHz, 160 kbps4,388,816,051 bytes (4.39 GB)
B. 10 min MOVH.264, 1920x1080AAC, stereo, 48 kHz, 128 kbps728,935,720 bytes
C. 10 min MKVH.264, 1920x1080Track 1 named System audio: AAC stereo, default. Track 2 named Microphone: Opus mono736,844,134 bytes
D. 10 min WebMVP8, 1280x720Opus, stereo, 48 kHz, 96 kbps requested81,204,433 bytes
E. 1 min MP4H.264, 1920x1080None71,877,497 bytes
F. 30 s MKVH.264, 640x360AAC, 5.1 (six channels), 48 kHz7,999,438 bytes

The built site was served locally and driven in Chrome 153.0.8010.53 on Windows 11, on an Intel Core i5-13600KF with 32 GB of memory, using Playwright. Advertising and analytics requests were blocked so the test did not count as visits. Every saved file was read back with ffprobe to confirm codec, sample rate, channels and duration. Run date: 5 October 2026.

What the File check panel reported

The panel appears under the editor once a file opens. Its bitrate is measured from the track's packets, because containers rarely store one.

FilePanel summaryReady to edit after
AMP4, 1 audio track, 1 video track. AAC, stereo, 48 kHz, 160 kbps12.0 s
BQuickTime File Format, 1 audio, 1 video. AAC, stereo, 48 kHz, 128 kbps1.3 s
CMatroska, 2 audio, 1 video. 1. System audio, eng, AAC, stereo, 160 kbps, default. 2. Microphone, eng, Opus, mono, 73 kbps1.8 s
DWebM, 1 audio, 1 video. Opus, stereo, 48 kHz, 69 kbps1.8 s
ERefused at open: This file has no audio track, so there is nothing to extract.0.1 s
FMatroska, 1 audio, 1 video. AAC, 6 ch, 48 kHz, 186 kbps0.2 s

Ready means the waveform for the whole timeline was drawn. That time follows the length of the audio, not the size of the file: the 4.39 GB hour took 12 seconds, because the picture is never decoded. File D shows why the bitrate is measured. FFmpeg was asked for 96 kbps Opus, but Opus varies its rate with the signal, and the simple test sound needed only 69 kbps.

Export times and sizes

FileOutputTimeSaved sizePanel estimate
A (60 min)MP361.4 s86,400,744 bytesAbout 86 MB
AM4A, unchanged copy6.1 s72,676,105 bytesAbout 72 MB
AWAV29.4 s691,200,044 bytesAbout 691 MB
B (10 min)MP38.8 s14,400,744 bytesAbout 14 MB
BM4A, unchanged copy1.2 s9,739,957 bytesAbout 9.6 MB
BWAV5.0 s115,200,044 bytesAbout 115 MB
C (10 min, track 1)M4A, unchanged copy1.3 s12,263,586 bytesAbout 12 MB
D (10 min)MP39.6 s14,401,320 bytesAbout 14 MB
DM4A, AAC encode3.9 s14,518,706 bytesAbout 14 MB
F (30 s, 5.1)MP3, folded to stereo0.3 s721,320 bytesAbout 721 KB
FWAV, six channels kept0.3 s17,292,140 bytesAbout 17 MB

Every estimate landed within 2 percent of the saved file. On this machine MP3 ran at roughly 60 times real time, so an hour of audio took about a minute. The unchanged M4A copy was about ten times faster again, because nothing is decoded or encoded. WAV sat between the two. A phone will be slower; these are desktop figures.

What ffprobe found in the outputs

Two cases the test checks

An unchanged M4A copy from an MP4 or MOV keeps the 1,024 samples of AAC encoder priming at the start, which the source hides with an edit list. The first packet therefore has a negative timestamp (-0.0213 s at 48 kHz), and the copy shifts every packet by that offset without changing its bytes. That is why each copy above runs 21.3 ms longer than its source.

A 5.1 track cannot go into MP3, which holds at most two channels. MP3 export folds it to stereo and says so before Convert is pressed; WAV keeps all six channels, as file F shows.

Reading these numbers for your own recording

For an hour-long lecture or meeting saved as MP4 or MOV with AAC audio, the unchanged M4A copy is the quickest and smallest: about 72 MB at 160 kbps, with no generation loss. MP3 at 192 kbps costs 86.4 MB per hour and plays almost everywhere. WAV costs 691 MB per hour for 48 kHz stereo and is only worth it when the next step is an audio editor. If the File check shows a recording with no audio track, the sound was not captured; no converter can recover it. For opening problems, the troubleshooting section of the format guide lists what each message means.

Advertisement

Questions about screen recordings

Is a screen recording an MP3?
No. It is a video file. The converter extracts its audio track to MP3, M4A or WAV and leaves the picture behind.
Is there a guaranteed maximum file size?
No. The page imposes no fixed cap, but completion depends on the browser's decoder, free local storage, device resources and whether disk-backed output is available.
Can it isolate one speaker?
Only if that source was recorded as its own audio track. It cannot separate voices that are already mixed together.
Why generated files instead of real recordings?
Generated files can be described exactly and rebuilt with the same settings, so a later run measures the tool rather than a different source. They match common container and track layouts, but a real recording may use other settings, and a speech recording compresses differently from a test tone.
Will my phone be this fast?
No. These are desktop figures. Phones decode and encode more slowly, and they stop background tabs, so keep the tab open and the screen on during a long export.
Was any file uploaded during the test?
No. The site ran from a local server and the files were read from the same disk. Advertising and analytics requests were blocked for the run.

Last updated 2026-11-20