MP3 to24bit: how bad can it get?

[hooty2]

Hi all... i grasp it's not ideal, but i'm wanting to know more clearly why not. An internet search hasn't clarified.
A friend made location recordings in a forest involving flutes and woodwind, looped guitars and various fx at source.
He hasn't been clear what the original files were recorded to but he sent me MP3's for an opinion and some ideas for improvement which he can feedback to the guy mixing etc. I listened and gave my three penneth worth and then dropped the MP3 into DAW/Reaper - 44.1/24 bit.
I wanted to see how much the original could be improved & maybe try out the vailidity of some of my suggestions, with little concern at this point of the quality ....so, and using the MP3 stereo file, i autometed EQ's to unmask a generally congested mix and bring a bit more interest, and bandpassed some elements onto new tracks....I didn't use any plugins except reverb and widening on the master track.
What is the technical side of Reaper (any DAW) making of the mp3 into 24 bit etc and how does the render at 24bit for soundcloud appraisal stack up to my usual in and out at 44.1/24bit throughout.
cheers

[James Perrett]

The one piece of information that you haven't given us is the bit rate of the mp3. This is important - if it is around 220kb/s or higher then it is hard to differentiate that from an uncompressed file without very careful listening. However, multiple encoding/decoding passes with different lossy compression algorithms will result in noticeable degradation, even if high bit rates have been used.

If you want to send them something to listen to then maybe send them a flac file via one of the cloud services - Soundcloud noticeably degrades the sound in my experience.

[SafeandSound Mastering]

Your MP3 has been created by a lossy encoder. So this involves complex maths and a lot of filtering. Information is now missing.

You have subjectively improved the sound and render to 24 bit lossless wav/aiff etc.

The original losses are baked into that 24 bit lossless container.

Then you will be cascading alogorhythms by uploading to SC, Premium users get 256kbps and non subs users I think it might even be 96kbps or 64kbps Opus now.

It can in some instances be LPFing the audio.

How bad is it, either use something that in real time encodes such formats (ADAPTR Streamliner / Sonnox) or encode it offline outside your DAW as something similar as possible and listen.

It may be reasonable or it might be dull with swirling modulated filter type sounds in the high end.

Give it a whirl and find out. I think this is what you are asking.

[Hugh Robjohns]

A friend made location recordings in a forest involving flutes and woodwind, looped guitars and various fx at source. He hasn't been clear what the original files were recorded to but he sent me MP3's for an opinion...

It's unlikely the source recordings were made as mp3 files. While most location recorders can do that, it would be extreme muppetry to have done so for something intended for release, especially when most recorders default to 24 bit or floating point, and have done for quite a while now.

What is the technical side of Reaper (any DAW) making of the mp3 into 24 bit etc...

It replays the mp3 file through a standard decoder to rebuild a close approximation of the original source waveform. However, the whole raison d'être of mp3 is to throw away elements of the audio the algorithms think a normal person, listening at a normal volume, can't hear, in order to reduce the file size.

Those elements, once gone, can't be brought back, and will include signals that the algorithms think have been frequency-masked by other, louder elements, for example, as well as low-level noises and extreme high frequencies... amongst other things. And all on a dynamic basis

One of the biggest problems with MP3s in the context of what you're doing, though, is that they were designed to be used only for the final delivery of a mix. You mix something, you encode it, your client plays it back and listens to it. That's it. But if, instead, you re-encode a lossy file the processing artefacts quickly start to stack up and become very obvious and unpleasant.

You hear this occasionally on BBC Radio (despite my best efforts all those years ago). It used to be a real bug-bear on the Steve Wright show where he often played audio snippets he'd borrowed from the interweb which were encoded in various data-reduced formats. They'd all sound acceptably fine in the studio, because he was acting as end listener (exactly as intended)... but by re-broadcasting them unto the Nation, that audio then went through further stages of lossy data reduction — most notably for DAB and BBC Sounds listeners — and so the real end-user audience would hear nasty artefacts which he (and his technical assistants in the studio) couldn't hear!

So, the absolute golden rule us to NEVER re-process an MP3 file... but if you do, don't ever do it if the resulting new file is likely to be re-encoded in a lossy format at some future point (especially in the same lossy format). And that is pretty much guaranteed to happen somewhere, sometime, be it on YouTube or SoundCloud, or many streaming services, or whatever.

... and how does the render at 24bit for soundcloud appraisal stack up to my usual in and out at 44.1/24bit throughout.

Try it and listen (carefully).... but I'd expect the encoding artefacts to become more obvious — although it can take good monitoring and some ear training to be able to recognise low-level encoding artefacts for what they are. Most listeners just perceive the result as being of poor quality.

The fact that you've mixed in some reverb might well help to mitigate the problems because that will have filled in some of the missing data filter bands and given the re-encoding process something new to chew on. But it's definitely not a fix!

So, in all seriousness, having completed your proof of concept, I'd strongly recommend you get hold of the 24bit (or floating point) source files and work properly with those. You'll probably find your EQ moves to unmask and declutter work a lot better and easier too!

[Drew Stephenson]

Downloadable soundcloud files are in the same format and bit-depth they were uploaded. Whilst the web-player may introduce all manner of nastiness the core file is not corrupted.

[hooty2]

Ask and you receive!... I am so grateful for this very detailed response from everybody and some pragmatic advice on how to proceed.
I'm texting on my phone so I'm not going to be exhaustive in this thanks but I'm really really grateful thank you

[The Elf]

Once an audio file has been reduced to an MP3 ideally it should be left there. You can never get the missing data back, and re-encoding it to any format compromises it further.

I've told the story here a few times of the band I worked with that complained the mixes I'd given them sounded bad... It transpired that each band member had been ripping from CD to MP3, writing back to a CD, ripping to MP3, writing to a CD... By the time it reached the drummer it sounded like it was playing in a bread bin!

[ef37a]

Could the OP not simply re send the files as uncompressed 24bit .wavs using wetransfer.com?

I am sure I have read here that it is error free?

Dave

[hooty2]

to tidy up:
i have asked the recording guy to send me files or a stereo mix on wetransfer. he uses a zoom recorder and does record at 44.1/24bit. he sent the mp3 as a quick means of appraisal (?).... probably didn't expect me to fiddle etc.
i now have the raw files and everythings fine between us.
However, my original question was valid and i remain grateful for the insights and education with clarity.

[Hugh Robjohns]