How each row is calculated
MP3 and AAC at a fixed bitrate grow by the bitrate divided by eight, in bytes, every second. At 192 kbps that is 24,000 bytes a second, 1.44 MB a minute and 86.4 MB an hour. WAV stores every sample: sample rate × channels × 2 bytes for 16-bit audio, plus a 44-byte header. At 48 kHz stereo that is 192,000 bytes a second, eight times the 192 kbps MP3 rate.
The unchanged M4A copy is different. It keeps the source track's own bitrate, so its size depends on how the recording was made, not on PrivateMP3. Type that bitrate into the calculator. The File check panel shows it for your file after it opens, measured from the track's packets rather than read from a label.
Checked against saved files
The formulas were compared with files saved by the tool in the screen recording test and read back with ffprobe.
| Output | Formula | Saved file | Difference |
|---|---|---|---|
| MP3, 10:00 at 192 kbps | 14,400,000 bytes | 14,400,744 bytes | +744 bytes (0.005%) |
| MP3, 1:00:00 at 192 kbps | 86,400,000 bytes | 86,400,744 bytes | +744 bytes |
| WAV, 10:00, 48 kHz stereo | 115,200,000 + 44 | 115,200,044 bytes | none |
| WAV, 1:00:00, 48 kHz stereo | 691,200,000 + 44 | 691,200,044 bytes | none |
| M4A copy, 1:00:00, AAC measured at 159,937 bit/s | 71,971,650 bytes | 72,676,105 bytes | +704,455 bytes (1.0%) |
The MP3 overhead is a fixed few hundred bytes of header, whatever the length. The M4A overhead grows with length because an MP4 file indexes every audio packet: an hour of 48 kHz AAC has 168,750 packets of 1,024 samples, and four bytes per packet in the size table alone is 675,000 bytes. The calculator therefore adds 188 bytes a second to both M4A rows for that index (187.5 at 48 kHz, rounded up). With it, the formula for the hour-long copy is 72,646,650 bytes against 72,676,105 saved, 0.04 percent short; without it the M4A rows could promise a fit that misses a limit by about 1 percent.
Worked example: an hour of lecture into a 25 MB email
Google gives the attachment limit for personal Gmail accounts as 25 MB; above it, Gmail replaces the attachment with a Google Drive link. (Gmail Help: attachment size limits)
At 192 kbps, 25 MB holds 17 minutes 21 seconds, so a one-hour MP3 from this site will not fit, and there is no lower bitrate here to change that. Two honest options remain. Send the file as a link instead. Or split it: open the recording, select the first 15 minutes, press Keep only selection and convert, then repeat for the next stretch. Each 15-minute part at 192 kbps is 21.6 MB, under the limit with room for the message. If the recording's AAC track is 128 kbps or lower and you need the whole hour in one file, the unchanged M4A copy is about 58.3 MB at 128 kbps, index included: smaller than the MP3, but still over the limit.
Megabytes are counted two ways
This site, the File check panel and most storage makers count 1 MB as 1,000,000 bytes. Some operating systems count 1,048,576 bytes and still write MB. The same 25,000,000-byte file is then shown as about 23.8. When a limit is close, compare bytes, not the label.
When the calculator will be wrong
- Variable bitrate sources. Opus and some AAC encoders change their rate with the sound. In the test, a WebM requested at 96 kbps Opus measured 69 kbps. Use the File check figure, not the setting you remember.
- Cuts. Removed time is not written. Enter the duration that will remain, which the editor shows as Output.
- Channels. MP3 and AAC at 192 kbps are the same size in mono and stereo; WAV doubles from mono to stereo. A 5.1 track exported as WAV keeps six channels and is three times the stereo size.
Measured: the source does not move the MP3 size
Every MP3 here is encoded at 192 kbps; there is no bitrate picker.
These are MP3 files the converter saved in desktop Chrome between 13 and 17 October 2026, from generated test recordings, set against bitrate ÷ 8 × seconds. They show the same thing every time: the saved file is the arithmetic plus a small fixed amount, not a percentage.
| Source track | Output length | Arithmetic at 192 kbps | Saved MP3 | Difference |
|---|---|---|---|---|
| AAC stereo 128 kbps | 10:00 | 14,400,000 bytes | 14,400,744 bytes | +744 |
| Same, two breaks removed | 7:30 | 10,800,000 bytes | 10,800,744 bytes | +744 |
| Same, part 2 of a 5 MB split | 3:28 | 4,992,000 bytes | 4,992,936 bytes | +936 |
| Same, part 3 of that split | 3:04 | 4,416,000 bytes | 4,416,936 bytes | +936 |
| AAC stereo 96 kbps (sample recording) | 0:20 | 480,000 bytes | 481,704 bytes | +1,704 |
| Opus mono (sample recording) | 0:20 | 480,000 bytes | 481,128 bytes | +1,128 |
Two things follow. For a size limit, the formula is safe to plan with: a part sized from it came out under 5,000,000 bytes with about a kilobyte to spare, which is how the size-limit advice in File check sizes its parts. And the source does not move the MP3 size: a 128 kbps or 96 kbps stereo track and a mono track all produce 24,000 bytes per second plus that small remainder, because the encoder target is fixed.
An unchanged M4A copy is different, because it keeps the source rate. For the same 10-minute 128 kbps track the copy saved as 9,739,957 bytes. The source rate gives 9,600,000 bytes; the calculator adds about 188 bytes per second for the MP4 packet index, 112,800 bytes here, which brings the estimate to 9,712,800, 0.3 percent under the saved file. Removing part of that recording turns the copy into a 192 kbps encode, which is why cutting it can make the file larger; the guide to smaller lecture audio has that measurement.
Bitrate cannot repair the source
A higher output bitrate cannot restore clipped peaks, remove room noise or reconstruct detail discarded by an earlier lossy encode. If a video contains AAC, converting that track to MP3 first decodes the AAC and then performs another lossy encoding. If the input is already MP3, cutting and exporting MP3 creates another MP3 generation.
There is no universal bitrate at which speech or music becomes indistinguishable. The result depends on the source, encoder, audio material, listening equipment and listener. Claims that 128 kbps is always transparent, or that nobody can hear a difference above 192 kbps, are not reliable operating rules.
M4A: look for the copy note before changing anything
M4A is an audio form of the MP4 container family; it is not another name for the AAC codec. PrivateMP3 can write M4A in two ways: copy the existing compressed packets into a new container, or decode the selected track and encode new AAC. The output extension is the same, but the quality history is not.
The packet-copy route is offered only when all of these are true:
- The selected audio track is identified as AAC, AC-3 or E-AC-3.
- M4A is the selected output.
- The complete audio timeline is kept.
- There are no cuts, volume changes or fades, and the clip task is not chosen.
When those conditions hold, the controls show Copied without re-encoding — same audio, no quality loss.
That visible note, rather than the M4A button alone, confirms the copy route.
Packet copy and AAC encoding are different jobs
Packet copy is also called remuxing: the encoded audio is moved into a newly written container without changing the compressed audio packets. The source video is not modified. The result contains one selected audio track; the video and subtitle tracks are not carried into the M4A.
A selection is not an edit until you apply it
Dragging over the waveform creates a selection for inspection. It does not by itself shorten the output. Pressing Remove selection or Keep only selection changes the kept timeline and disables packet copy. A volume adjustment also changes decoded sample values, so it requires re-encoding.
To preserve eligible AAC, use this order:
- Open the local video and choose the intended Audio track if more than one appears.
- Leave the complete timeline and volume unchanged.
- Select M4A and confirm that the copy note is visible.
- Convert and save the new audio-only file.
If you need a cut, make it deliberately and treat the result as newly encoded AAC. The old claim that a copied file could still be cut on an AAC frame boundary did not match this implementation: any applied cut switches to the decode-and-encode path.
Measured: what the copy keeps
These copies were made in Chrome on Windows 11 between 5 and 10 October 2026 from FFmpeg-generated files, and each saved M4A was read back with ffprobe. The method and the rest of the results are with the screen recording test.
| Source | Copy took | Saved M4A | What ffprobe found |
|---|---|---|---|
| 60 min MP4, 4.39 GB, AAC 160 kbps | 6.1 s | 72,676,105 bytes | AAC, 48 kHz stereo, 159.9 kbps, 3,600.021 s |
| 10 min MOV, AAC 128 kbps | 1.2 s | 9,739,957 bytes | AAC, 48 kHz stereo, 128.3 kbps, 600.021 s |
| 10 min MKV, AAC 160 kbps on track 1 of 2 | 1.3 s | 12,263,586 bytes | AAC, 48 kHz stereo, 159.9 kbps |
| 30 s MKV, AAC 5.1 | 0.05 s | 710,176 bytes | AAC, 48 kHz, all 6 channels kept |
| 3 min M4A, AAC 219 kbps | 0.09 s | 4,965,348 bytes | AAC, 44.1 kHz stereo, 219.2 kbps |
In every case the bitrate, sample rate and channel count came out as they went in, which is what an unchanged copy should do. Two details show up in the numbers. Each copy runs 1,024 samples longer than the source, which is 21.3 ms at 48 kHz and 23.2 ms at 44.1 kHz: AAC begins with that much encoder priming, which the original hides, and the copy keeps it at the start rather than altering any packet. And the M4A is about 1 percent larger than bitrate times duration, because an M4A indexes every audio packet; the File check panel and the calculator above both include that index. A 5.1 track copies with all six channels, unlike an MP3, which can hold only two.
What happens when the source is not AAC
An M4A can still be produced from another audio codec when the current browser can both decode that track and encode AAC. PrivateMP3 then makes new 192 kbps AAC. This is a conversion, not a lossless container change. It may still be the right output when the destination requires M4A, but it does not preserve the original compressed packets.
If the selected track says no decoder, a newly encoded M4A cannot start. Try another audio track in the same file. Unchanged AAC is the exception because its packets can be read and written without decoding samples. PrivateMP3 does not claim that every codec combination works in every browser.
Copy failures have their own messages
The copy path can fail while reading compressed packets or while finishing the new M4A container. The compressed audio packets could not be copied from this file
points to packet reading and may indicate an incomplete source. The output file could not be finished
occurs later and tells you to check available storage; it does not mean the original was changed.
A missing track, unreadable duration or waveform failure happens before conversion and is a different problem. Use the troubleshooting section of the format guide to identify the stage instead of assuming that every M4A failure is a codec problem.
WAV: the extra bytes do not add detail
Video sound is often stored as lossy AAC, Opus or Vorbis. Converting it to WAV decodes what remains after that earlier compression; it cannot reconstruct missing detail or clean up the microphone. PrivateMP3 writes uncompressed 16-bit PCM. That gives an editor direct samples to work with, but the larger file is not a better recording.
The 16-bit detail matters with unusual source material. If a video contains PCM with more than 16 bits of precision, this WAV output reduces it to 16 bits. Most ordinary video soundtracks are already lossy, so their earlier codec is usually the larger quality constraint.
Use WAV as a working file
WAV is useful before editing in a digital audio workstation, signal analysis, speech cleanup or mixing. Those jobs commonly work with PCM internally, so WAV avoids adding an MP3 or AAC encode before the work starts. Make the smaller delivery file after the edits. For ordinary listening, choose MP3 when compatibility matters. If the source is AAC, AC-3 or E-AC-3 and you need the complete, unchanged track, M4A may copy it without re-encoding.
Measured: mono and stereo sources
Tested on 6 October 2026 in Chrome 154.0.8037.98 and Edge 154.0.4258.53 on Windows 11 (Intel Core i5-13600KF, 32 GB), with files generated by FFmpeg from a 220 Hz tone and quiet pink noise. Times are from Chrome; Edge differed by at most 0.7 s on any row. Each saved file was read back with ffprobe. The method is described in the screen recording test results.
| Source, 5:00 | WAV | Formula |
|---|---|---|
| MP4 with mono AAC, 44.1 kHz, 96 kbps | 0.8 s, 26,460,044 bytes, 1 channel | 44,100 × 1 × 2 × 300 = 26,460,000, plus a 44-byte header |
| MP4 with stereo AAC, 48 kHz, 128 kbps | 2.4 s, 57,600,044 bytes, 2 channels | 48,000 × 2 × 2 × 300 = 57,600,000, plus 44 |
The mono source stayed mono and kept 44.1 kHz, as the formula above says it will. For the same five minutes the mono file's other outputs were an MP3 of 7,201,174 bytes (mono at 192 kbps, in 3.4 s) and an unchanged M4A copy of 3,665,528 bytes (in 0.3 s). The WAV is about seven times the size of that copy and holds the same decoded sound.
Which of the three outputs the next application accepts is a separate question; the format guide starts from that.