Ok, I don't want the discussion to get lost in semantics, but I think it's important that we stick to facts.
What you can do in StereoTool is remove, reduce, enhance and generate information.
There are no processes that bring back something that is no longer there. And there are no processes that only reduce compression artifacts and leave the original data untouched. With every process, you generate new artifacts that weren't there before. The trick is to reduce as many compression artifacts as possible while doing as little damage as possible to what was originally contained. But damage always occurs. We just try to hide it.

 

Last edited: Jun 26, 2022

I don't think it is just a makeup. It uses mathematical models to get the most probable missing values from the existing (available) ones. And those models are based on known mp3 specification (it is exactly clear what procedures mp3 lossy encoding uses, and what are the consequences). It helps greatly to undo things (to a degree) in mathematically most appropriate way [it is a sort of a guide to undoing the lossy encoding]. So it has noting to do with individual taste or embellishing.
Hence, it is just matter of how close those approximations are to the original state. And they might be so close that while listening you do not have a remotest idea that some restored file originates from a mp3 file.
And mp3s are tiresome to listen, indeed.

 

Last edited by a moderator: Jun 26, 2022

I don't know where this much idealism in this thread comes from. You can not do a single gain change to a file and then undo it (after a save) without data loss, because of digital rounding. You can not completely (if at all) undo any executed audio DSP action.
Nobody here ever claimed that there is a way to completely restore a lossy file (like 'backup and restore'). Refuting this is just a waste of time. Never in my life it occurred to me that such thing is possible, and I don't know why those refutations are addressed at me at all.

 

All in all, your comments are well-founded, correct and comprehensible. I just wanted to point out that some excerpts contain minor flaws that might confuse some readers with less experience and lead to wrong conclusions.

 

Did you ever try dBpoweramp Music Converter its been around for over 20 years and is really great, easy to use and has a codec database if you need additional codecs for it. Check out the website for info.

 

This

I guess the whole discussion of restoring old mp3's is dedicated to private recordings, or, in any case, audio that cannot be found in other ways.
As far as I'm concerned, I completely abandoned the concept of lossy codec from my listening, my ears don't deserve it.
It's been a while since I've been looking for all the digital catalog I already own in loseless format: Searching for tens of thousands of files is a long and tedious job, but in the end I avoid getting pissed off when I listen to music. When I intercept on youtube, spotify, bandcamp, etc. a new song that interests me, I immediately look for it in its uncompressed version.
Apart from the usual 44/16 flac, many distributors also offer better formats such as 48/24 and even 96khz, very interesting is also the DSD.
Technology goes on: no, we won't miss mp3, I really don't think it can become a "vintage" sound in the future, (like the vinyl saga): mp3 is now just an outdated format

 

Last edited by a moderator: Jun 27, 2022

Nowadays when bluetooth gains more and more popularity, mp3 appears not that guilty.
The horrible compression in the mixes and everything else are influenced by these sad directions.
Why would you request lossless format for the new records that are meant to reign in BT earbuds?

 

There are some obstacles to this approach.
1. many valuable songs you just can not find on the internet other than in a lossy format, and some even only in low bit-rates. This is valid especially for non-english recordings, and it may be drastic (for Greek music my estimate is about 80%).
2. many lossless songs across the internet are fake, meaning FLAC containing a lossy source.
- One should first thoroughly check his lossless collection for fakes. I even found some commercial audio CDs containing lossy sources. In my case, more than 15% 'lossless' files on my PC were fakes. Now I always first test the FLACs after downloading, if it is a fake I search for another one, etc.
And there are also even more mp3 fakes (lower bitrate files re-encoded to a higher one).

 

Last edited by a moderator: Jun 27, 2022

Thanks for all the inputs.
My original question was not about "restauration". I know that everything gone is lost.
It was about "enhancement", just like a mastering suite do on "basic" songs to "enhance" them ... but with a special threatment for mp3 compression.
Will try Thimeo, it looks pretty usefull.

 

I agree about the above written.

Do you think this song could be improved with that tool you mentioned?

 

Passing by and just wanted to say I like your new signature @Sinus Well

"I am a meta-sexual pseudo-TrIPoC & non-binary conspiracy mystic unicorn. My pronouns are Fxszrt/Fxsczprt and tcxzfgl/tcxzfglkx"


Maye I'll use that next time I use Tinder... almost beats my draft til today:
"I'm just an outdated boomer-white-hetero-cisgender-omnivore. And for that, I'm sorry. Pls don't report me"

 

Last edited: Jun 27, 2022

Of course, it is just another audio file, like the rest of them.
Do you have it in mp3 (not including 'mp3' downloaded from YouTube)? Post it and I will restore it for you.

 

I must admit, I can see the potential but it is a little intimidating. I started with an 128kbps mp3 that's clearly clipped - lot's of lopped off peaks at 0db. To begin begin with, I thought I'd concentrate on the De-clipping aspects of this software but it's been a tough start. I tired the defaults in Basic mode and although some of the peaks were slightly declipped, the majority were still flat lines at 0db. I figured out that in order to get to the more advanced settings I needed to change the Operating Mode to something more suitable but this revealed so many additional parameters that I didn't know where to start and the help within the app isn't great either.

So, I suppose my question is, if declipping in default mode isn't 'aggressive' enough to smooth the clipped parts of the waveform, which parameters would you start with in order to 'increase' the declipping capabilities of this software to the point where the clipping is removed? I appreciate this is probably a difficult question to answer but I did wonder if you have a process or work flow that you apply to such scenarios in order to find the optimum settings?

I also tried RX 9 and the Sound Forge declipper. I was quickly able to get good results with both of these however I can't help feeling that mastering this software would provide more flexibility and also it'd provide a one stop shop for all the otfer processing needed to 'refresh' mp3/lossy files. I'm prepared to put in the time to learn this but it's hard to get started due to the sheer number of tweakable options.

Any thoughts would be very much appreciated.

Or, are there any good tutorials out there that deal with this?

Many thanks

 

Professional tools need some months to master.

 

Last edited by a moderator: Jun 27, 2022

I have it from the Napster era. Since that time I haven't found it for 2 decades, just these days discovered the name of the album and of course it's not on Spotify (which would also be just 256 bit lossy format anyway).
Actually I think my file may be in a bit better quality than the youtube one, I just posted that link first because I thought it's easier.
Now that I found out on which album was this song, I can purchase it in the future.
But I'm curious about how much improvement these software can achieve. Not recovering but recreating the sound I guess.
...Oh sh*t, I can't post zippy link here. I'll send it to you in PM.

 

https://www.discogs.com/release/3939325-Szakcsi-Straight-Ahead

 

Just rar the file first and then send it via zippyshere. I can not download anything from your shortlink, it just offers me to shorten 'my link'.

 

The file is re-encoded from 128 to 160 (fake bit rate and double filtering, compression, etc.) Not good for a start.
There is only a single clipping in the whole file.

 

Now here it is. To hear the real difference you have to gain match the mp3 with mp3Gain and listen on some decent stereo system.
https://www64.zippyshare.com/v/Ywtm3Ipg/file.html

 

I can't argue with that! I think I'm struggling a little with Stereotool (ST) because the documentation is quite scant and there aren't many video tutorials like there are with some of the more popular plugins/tools. Having said that, I am making good progress.

I've also figured out one of the more fundamental issues I was having with an mp3 file that had obvious clipping at 0db. I was able to get good results with other tools (even the predecessor to ST - Perfect Declipper) but I just couldn't get the declipping aspect of ST to work in this scenario - it simply would not correct/rebuild these clipped parts of the mp3. It then dawned on me that the ST declipper needed some headroom to rebuild the clipped audio. I checked Perfect Declipper and saw that, by default, it attenuates the signal by -6db thus providing headroom for the declipper to do its thing. ST doesn't do this (by default).

Back to ST, I tried using the Pre Amplifier setting in the Processing settings to attenuate the signal and this then allowed the declipper to handle the clipping that was occurring at 0db (prior to attenuation) and the clipped part of the signal was rebuilt . However, the help associated with the ST Pre Amplifier setting confused me. It states "Declipper, Link error '10182' and AGC if Gating based on volume before Pre Amp is enabled ignore this setting. If you want to correct a too low sound card input level, use Input gain instead." This seems to suggest that Declipper ignores this setting. It also suggests that I should instead use the Input gain setting however this setting only allows amplification of the signal, not attenuation.

So, my question is, if an mp3 has clipping at 0db, how do I handle this within ST? Do I use settings within ST to attenuate the signal thus providing headroom to allow the declipper to rebuild the signal that was lost due to clipping at 0db OR is it good practice to attenuate such mp3s PRIOR to processing them with ST's declipper?

Forgive me if this is a silly question but I am on a learning curve and I'm not sure what the best approach is for this scenario. Any help is most welcome - thanks