I'm looking for a MP3 enhancer plugin to convert 128k mp3 to wave.
I know that nothing can recreate frequencies removed by compression, but maybe some plugs can do the trick ??
Any idea ?

 

Hi, Zynaptiq UNCHIRP was designed to do this.

Also, I seem to recall that Acoustica Digital Audio Editor has a plugin called Vitalize which may do this kind of thing too.

 

Last edited: Jun 24, 2022

put acustica audio on it - the breath of life!

 

RX has a function to recover lost frequencies, some time ago I saw a video that showed how to repair low resolution mp3.
Here I found one that does it with the voice, but I imagine it's the same.

Edit:


I remembered another tool, because I remembered this video. the vst is called Zynaptiq Unfilter . But like Recycle said. The best option would be to find the Flac. But if there is no way, I think this tool can "help" look what he does with this song

 

Last edited by a moderator: Jun 24, 2022

I'm afraid no plugin will do the job decently. I tried Izotope RX: well... consider it as the last resort. In case your file comes from already published material (basically a song), the best option is to find the flac file of it

 

You have everything you need in Thimeo Stereo Tool. In fact, its basic function is recovering mp3 files.
And it's also top software, the real professional high end. It is manly aimed at radio stations, as a standalone program,
One can see more on Omnia presentation (they use Thimeo license):

but it is available in VST2 and VST3 (even VST2 x86).
It consists of many modules, and each module has a plethora of settings. Study them carefully. Use Configuration/User Interface/Operating mode/Expert.
The basic modules you should use are [repair section] Declipper, Dequantizer and Delossifier. [processing section] Natural Dynamics and Absolute Highs. [processing/stereo] Azimuth (for correcting the mp3 stereo image artefacts).

P.S. If you use Thimeo, don't use any other software that tries to do anything similar to its functions.

 

Last edited by a moderator: Jun 24, 2022

I tried Stereo Tool: listen, it's not good at all, it keeps adding messes on an already damaged file. it's like putting barbecue sauce on top of a dog shit. Maybe it's my fault that I don't get the settings right.

ok, it's time for the
Stereo Tool Challenge of the day:
prove me that it is possible to revive this bad mp3 encoding

This was a wav file encoded (and destroyed) with a 96kbps mp3 compression: can you make it sound decently with Stereo Tool plugins?


-------

Reference wav file:

 

Last edited by a moderator: Jun 25, 2022

First, I don't know how to download this. I get it automatically opened in a player, with no download option.
Then, testing at very low bitrate will not reveal the power of Thimeo Stereo Tool. Since it works by using extrapolation, it needs some data to work with. mp3@96 gives a compression ratio of 14.7:1. It is unreasonable to expect to do anything much with such crap.
With mp3@192 and above after using Thimeo Stereo Tool you can hardly tell the restored file from the original.
You can use 128 and 160kbps files, but there will remain some audible differences, although the result will be convincingly better than the original.
If I understood you well, you expect me to restore your file and then post it here?
If so, post a link for download.

 

Click on the frame then on the appearing arrow on the right side, from there you can download it.

 

I skipped a few minor steps, but basically this is it.
https://www92.zippyshare.com/v/TNlURoZB/file.html

 

iZotope RX could help, Stereo Tool is a good suggestion too, but it's more advanced and harder to get quick results.
Unmask, TEOTE, BBE Sonic Maximizer etc. might work.
Essentially everything above 16K is gone so you'll need something to synthesize high end, but there's a decent chance you might not be able to hear well above 16K so don't overdo it.

 

Thank you for taking the challenge and spending your time in this test. The point was to understand if a skilled user like you could perform the miracle using Stereo Tool wisely (because I have tried it and I have not been able to come up with anything good).
Listening to the restored file: yes, I hear enhancing of the audible hi-end frequencies already present on the mp3 file. Unfortunately it is difficult to emphasize non-existent frequencies (that mp3 has a dramatic cut above 15k). Furthermore, the typical "chirp" of the encoder is still there and the transients remain chewed as expected.

A visual representation of the 3 files:

The real thing:



The 96kbps mp3:



Pratyahara restored:


I'm sure everything possible has been done to restore the file, but my conclusion is that you speak too enthusiastically about the app: the miracle just isn't there. based on what you said I was ready to be shocked, I was hoping for a new technological advance in audio production ...
We will probably have to wait for an AI app that smartly intervenes on the damaged file trying to "imagine" what the uncompressed original was like.

 

Last edited by a moderator: Jun 26, 2022

Very interesting thread. But my question is:
without reference file can you "restore" information that you don't know?
I mean, you can for sure try to make the audio more pleasant and remove to some degree annoying artifacts, but that is colorize.
It resembles me the bs in the the movie No Way Out (1987) where from an handful of pixels they recover a detailed picture of the spy.
So, the answer: no.
I agree that recover depend much on the quality of lossy compression; the less compressed, the better is.
But after a knee point (96kbps?) recover becomes a customization, even if well performed.

 

As I already said, the point of Thimeo Stereo Tool is not to 'restore' bottomless files, but to bring decent mp3 files closer to the original.
Although I warned you, I agreed to do this inadequate task just because you insisted. Otherwise, I would never do such a dead-end thing.
Every tool has its domain, and this was beyond it.
P.S. In restored file's graph, the highs reach 18,250Hz, and you can't hear them?
The chirp is result of extreme degree of quantization @96kbps, and I doubt any AI could undo it.
And the transients are much improved, compared to the original mp3@96kbps.
Nobody promised 'miracles'', but just substantial audible improvement, and that is what happened.

 

Last edited by a moderator: Jun 26, 2022

The premise is false. You can't "recover" anything here. Mp3 is a lossy format. What is lost is lost.
All that can be done is to reduce or mask unwanted artifacts by extracting information from the spectrum with supressive filtering or by padding it with harmonics or delay lines.
Dry mono signals can usually be processed relatively well in this way if the mp3 compression was performed on a lossless source signal. If a lossy signal is compressed again with the codec, an acceptable result is hardly possible. Moreover, the more L-R signal is present in the original, the more difficult it is to achieve an acceptable result.

Or in short: Voice recordings that are lossy due to data compression can be subjectively "improved", more complex signals like full-range music rather not.

 

To 'recover' does not mean to reverse the lossy file to the bit-exact copy of the original, but rather to remove some of the artifacts, and restore some of the lost data.
What Thimeo can do is 1. declip (best in the world, with 5 algorithms of analysis), 2. selectively multiband expand the compression (of course some data are irreversibly lost), 3. de-quantize (as far as can be done by extrapolation), 4. remove pre-ringing, 5. compensate for some psychoacoustic filtering done by lossy encoding standards, 6. extrapolate some frequencies lost by filtering by wave shaping, 7. correct stereo image errors, etc.
Is this not substantial? Especially when summed up. To some people this is negligible...?
And no other software mentioned in this thread could do 10% of this (not to mention the quality).
Anyway, for people who do not respect this tool, there is always an option to stick on to their crippled mp3 files.

 

I work with thimeo and similar tools all the time. I therefore know what it can do. I do not deny that it is tremendously useful in restoration. But what it can't do is recover lost data.

 

You can simulate the lost data, so pretty much nobody could tell the difference.

 

it's like Healing Brush in Photoshop. You can have a pleasant result (with multiple repeated attempts and needed skills) but it will always be your new version of something.
Restoration on archive audios that have an important content (like an important person's speach) to make it more audible is one thing. Recreating something that has been lost is another.
Better to thing this way: you want to remove (or reduce) artifacts and colorize the audio with your taste.
Don't think about recovering lost frequencies. (agreeing with the OP's optinion)