generation quality loss in MP3

I guess it’s not a new topic, but I couldn’t find it in the topic list. I think I need a simple yes/no answer. Stripping off the tech details: if I take a MP3 file at a given quality level (bit rate, etc) and re-export it at the same quality, I still get resultant quality loss due to the additional generation?
The answer will determine how I manipulate my music…

Yes. There is some quality loss every time that it is (re-)encoded to MP3.

The loss can be minimized by using a high bit-rate (at the expense of a larger file size).

[u]MP3DirectCut[/u] can do some limited editing/processing without decompressing.

[u]Nine different codecs 100-pass recompression test[/u]

There are editors that can work directly on the MP3 without all the format shenanigans. See: DVDdoug. They can allow some editing and still keep the original compression quality.

They’re very limited. Cutting, slicing, and deleting are good, special effects, filters, and volume changes as a rule are not.


Got my reply and learned my lesson. Thanks, guys.

I need cutting and slicing, Koz, but - unfortunately - volume change is a must.

Thanks again


Actually, MP3DirectCut can change the volume (including normalizing and “fading” in or out). But a limitation of MP3 is that volume adjustments can only be done in 1.5dB steps.

If you are “loudness matching” MP3gain also works losslessly (in 1.5dB steps.)

…I’ve only used MP3directCut a few times so I probably can’t help you with it.

…Sometimes you don’t have a choice, plus you might not hear any quality loss. But it’s something you should be aware of and you’ll know to minimize the number of times it’s re-compressed.

Thanks, Doug. I think that thanks to the answers I got here, I’m all set to optimize my music quality. Please, tell me if I am wrong: (1) I get my music at highest quality I can, and in WAV format (quality depends on the source: best from old CDs, not so best when I record from YouTube (I use Audacity recorder, and record as WAV) (2) I do all the trimming, adding silence, etc without exporting (3) I use Audacity to normalize/equate volume of perceived loudness, down to -18 LUFS (I think it’s Audacity’s default), no export yet (4) I amplify (with Audacity’s “effects” the volume to some higher level (usually I use amplification to 0 Db, so as to avoid clipping), and THEN I export the result into MP3. So I do all the work in WAV, and export at the end to MP3. Actually, since I do all in WAV, I can export stuff, then re-import it as WAV, as long as export to MP3 is done only once, at the end.

Am I wrong?

A different question: I listen to - and record - quite a lot of classical music. I am ASSUMING that I need higher quality MP3 for, say violins, than for piano (due to the higher frequency of the violin). However, I don’t know how the MP3 encider works. I tried to read about it in Wikipedia, and understood most of the words, but hardly any of the sentences, so I need someone who DOES understand the issue to confirm that my higher frequency-higher necessary quality is correct.

Many thanks

To be very nit-picky, a normal 16-bit WAV file is “almost” perfect.
Audacity uses an even more precise format internally, with twice as many bits per sample - “32-bit float”.

If you want a perfect round trip from export to import then you can export as “32-bit float WAV”.
In practical terms, the difference between export / import 16-bit WAV and export / import 32-bit WAV is usually too small to be worth bothering about.

The downside of 32-bit float WAV is that files are double the size of 16-bit WAV, and a lot of audio players can’t play 32-bit float WAV, but if you need perfection, then 32-bit float WAV is the answer.

For anything important, it is a very good idea to also export a backup copy in WAV or FLAC format. If, in the future, you ever need a better quality, or smaller MP3 file, then you can make a new MP3 from the backup.

A different question: I listen to - and record - quite a lot of classical music. I am ASSUMING that I need higher quality MP3 for, say violins, than for piano (due to the higher frequency of the violin). However, I don’t know how the MP3 encider works.

I don’t think violins are particularly difficult to compress but I’m really not sure. The high frequency loss from MP3 encoding is easy to measure but not always heard.* At higher bitrates/quality settings the most common audible artifact is something called “pre echo” that shows-up with transient sounds like castanets.

If you want MP3 but you are not too concerned with file size you can simply choose 320kbps constant bitrate (the “best” constant bitrate and the largest MP3 files) or VBR-0 (the “best” variable bitrate). I use VBR-0 for everything. We really can’t say higher bitrate is better unless it actually sounds better and that’s not always the case… Lower bitrates can often sound identical to the uncompressed original (in a blind ABX test) or if you hear a compression artifact at a lower bitrate it might also be present at higher bitrates.


  • In the context of music the highest frequencies are masked (drowned-out) by lower high-frequency sounds. Even if you can hear to 20kHz in a hearing test, your ear has less sensitivity at the extremes and those highest-harmonics in music tend to be weak. The main “trick” with MP3 is that it throws-away sounds that are masked. But it seems to throw-away a lot of those higher frequencies without a lot of analysis of what (if anything) is drowning them out.

"(quality depends on the source: best from old CDs, not so best when I record from YouTube (I use Audacity recorder, and record as WAV) "

I post recordings on youtube of jazz music, my channel is ctproduced.

The problem recording from Youtube is they have already applied a set of filters to the sound before your receive it to capture and record, so you pay an additional penalty, once from the original creation of the music/file uploaded, and a second time when they stream it to you.

I did a load of tests before creating the channel to decide if there was a “better” format to upload so that when it was heard, it would sound better. I’ve not redone those tests in two years, but I suspect they’ve changed something in the past 6-months.

I’ve added music to videos as .wav/32 files, and recorded what youtube streams back it’s definitely not the same and has been compressed. In the end, I pretty much universally use Audacity export to 220-260 kbps (best Quality) joint stereo and apply equalization to adjust the bass and treble slightly, leaving the midrange alone.

You can record this as a 32-bit .wav file, but you really only have a vbr .mp3 in quality. Even I add 32 .wav quality sound to a video, you don’t have the option to record it, since youtube have already culled it.

If I record what I get back, it’s pretty close. As far as I can tell, YouTube still isn’t doing any kind of Loudness normalization, and their levels are still all over the place. It’s not clear how they’ll resolve this, but eventually will have to.

Here is a short sample I posted recently, recorded in Audacity from virgin, first play vinyl.

I’ve just upgraded this morning to Audacity 3.0 to work with the new Loudness Normalization for the projects I post on mixcloud, also ctproduced.

Thanks for the info. I, too, use Audacity recorder and record as WAV from YouTube. Also, i - too - believe (or used to believe) that their streamed music are lowered in quality, but I reied analysing the sound I get visually (using Spek – Acoustic Spectrum Analyser - a nice little program), and found that, compared to the youtube music I download, recording the stream is much better. Listening to the results I get, I think I am pretty satisfied.

The trouble is that some of the music I save is old jazz, and you simply cannot get CD quality, unless you buy zillions of CDs, most of which contain 90% of stuff you don’t want, plus the one piece you are looking for, and purchasing high quality (sound-wise) MP3 or WAV piece by piece is impossible. The various seller simply don’t have the variety I am looking for, whereas YouTube does…


It depends on how the download is done. Websites that offer a download version will often produce better quality in the download than is available through streaming. YouTube does not provide downloads, so downloading is often via an online service that effectively streams the content to another server, converts it to a file (“records” it) and then sends it to you as a compressed audio file. When done this way the downloads are lower quality than the streamed audio.

Thanks. This explain the quality difference I got. I used Any Video Converter to download from YouTube (and then stripped the audio part with Audacity). Of course, I have no idea How AVC grab the clip, but I guess you’re - it’s some quality-reducing process.
I also tried YouTube premium membership, (which allows better audio), but found out I couldn’t use it on my PC (only tablet), and there was no way I found to record from the tablet.
At least you can’t blame me for not trying…