MP3 or AAC from a Video: Which Choice Keeps the Original Track

The short answer

For an MP4 or MOV with an AAC soundtrack, choose AAC in Audio from Video. The track then comes out as it was: in our test, all 1,408 audio packets matched the video's, byte for byte. MP3, the default choice, decodes the track and encodes it again at 192 kbps.

The copy was also the smaller file: 479 kB, against 704 kB for an MP3 of the same 30-second clip. Pick MP3 when the player or the form on the other end accepts nothing else. An AVI with MP3 sound works the other way round: there, MP3 is the choice that copies. A WebM with Opus sound has nothing to copy, whatever you pick.

How we measured

On October 22, 2026, we ran seven test videos through three pages on pulsarkit.com, in Chrome 155 on a 2023 Mac. They were Audio from Video, Video to MP3 and MP4 to MP3. The MP4 clips held 30 seconds of 1080p video and synthesized music, turned into AAC by Apple's encoder at 128 or 256 kbps.

One MP4 was repacked into a MOV, and into an MKV with a second AAC track of different sound, tagged English and Russian. The rest: a 10-second AVI with 128 kbps MP3, a 30-second WebM with the same music in Opus, a silent MP4 and a 59-minute MP4.

For each result we read the codec and bitrate, then compared its audio packets with the track inside the video, one checksum per packet. The clock ran from the button to Download, in a new tab each time. All three tools run in the browser tab, so the videos stayed on our computer.

Which choice copies the track

Audio from Video copies when the track already has the format you pick. The page then says so: “The aac track was copied straight out of the video without re-encoding, so nothing was lost.” Our MP4 and MOV gave the same 479 kB file. Each MKV track came out with its own packets intact, and the 256 kbps clip gave 949 kB.

From the AVI, MP3 copied the 128 kbps track: 157 kB, all 384 packets as they were. AAC re-encoded the same track into 238 kB at 192 kbps. The WebM's Opus track has no matching choice. MP3 and AAC both re-encoded it, to 704 kB and 729 kB, against 483 kB for the track itself.

Two settings under More settings change the outcome. Bitrate is ignored when the track is copied. Even out the loudness always re-encodes: with it on, AAC from our MP4 came out at 733 kB, and the copy message was gone.

Video to MP3 and MP4 to MP3 never copy. They encode with LAME every time, even when the video already holds MP3. From the AVI, Video to MP3 made a new 236 kB file at 192 kbps. If you are not sure what a video holds, try AAC in Audio from Video first. The copy message tells you whether you got the original.

Size, time and what re-encoding takes away

On a 30-second clip, every choice finished in 2 to 2.6 seconds, copy or not. Length shows the difference. Our 59-minute MP4 of 70 MB gave a 55 MB AAC copy in 2.2 seconds. The 192 kbps MP3 took 22.4 seconds and weighed 81 MB, more than the video itself, whose picture was tiny.

WAV is the third choice, and for these files it is never a copy: it is the decoded track, 5.5 MB for 30 seconds. Its samples matched our own decoding of the AAC track to within one step of 16-bit rounding. So WAV adds size but no second loss, which suits an editor.

To see what MP3 costs, we subtracted each file from the original recording: the quieter the remainder, the less was lost. The 128 kbps AAC copy left a remainder 14.6 dB below the music, and the 192 kbps MP3 made from it 14.1 dB. From the 256 kbps track, the copy left 19.5 dB and a 320 kbps MP3 17.6 dB.

That 320 kbps MP3 dropped the sound above about 20.5 kHz, while the AAC copy kept it up to about 22 kHz. It also weighed 1.1 MB, against 949 kB for the copy. We ran no listening test. How MP3 bitrates compare with each other is covered in MP4 to MP3 at 320 kbps.

The 44 milliseconds at the start of an .aac file

A copied AAC track lands in a bare .aac file, and that format has no room for one detail of the MP4. The encoder that made our track, Apple's, put 2,112 extra samples in front of the sound, and the MP4 tells players to skip them.

The .aac file keeps them. Our copy starts with 44 milliseconds of near-silence, its loudest sample at −42 dBFS. After that, every decoded sample matched the MP4's exactly. Played on its own, the file simply starts 44 ms later.

If you put the track back under the picture in an editor, cut those 44 ms first, or the sound runs that much behind. The WAV from the same page started in step with the video, sample for sample.

Videos with more than one audio track

Audio from Video takes the first audio track unless you set Audio track under More settings. It counts from one, in the order the tracks sit in the file. On our MKV, 1 gave the English track and 2 the Russian one, each packet for packet.

The result shows the file name and size, not the language or title of the track it took. Play the first seconds before you rely on it. Asking for track 3 of our two-track file returned track 1 without a warning. Video to MP3 has no track setting at all: it converted the first track.

A video with no sound at all gets no empty file. The page says “This video has no audio track.” and offers Open Add Audio to Video.