Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Merging two different encodes to improve quality (Read 2732 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Merging two different encodes to improve quality

Is there a way to do this properly? I have one encode which is a 44.1kHz 128kbps mp3 file and one which is a mono 24kHz adpcm. I know I'm not going to get a perfect result, but both of the files have certain advantages over one another and I really wish I could just merge them together without causing some kind of imbalance or artifact. Is it possible?

Re: Merging two different encodes to improve quality

Reply #1 –
Well, you can first extract the difference, and then add it to other file. For laughs. But I don't know if it will work that way, though.
Error 404; signature server not available.

Re: Merging two different encodes to improve quality

Reply #2 –
Thanks! I've already tried that, I just wanted to verify that it was valid and wasn't completely messing up the audio somehow. The result is a little interesting, and seems to vary a lot from file to file.

Re: Merging two different encodes to improve quality

Reply #3 –
Okay, I'm back again.

I just found out that there is another mono ATRAC3 version of the music, and attempted to align it with the mp3. Both of the files are 44100Hz, but it seems that there is some kind of difference in speed or something? I can't really explain what I'm hearing that well, so here's a video demonstrating it:

https://youtu.be/DzBKLKiCVmg
 

Re: Merging two different encodes to improve quality

Reply #4 –
Well, I would say that there is some kind of cyclic destructive and constructive interference, so I think there may be a tiny difference in speed as you suggested.

Re: Merging two different encodes to improve quality

Reply #5 –
Quote from: Ophy on
Well, I would say that there is some kind of cyclic destructive and constructive interference, so I think there may be a tiny difference in speed as you suggested.
Is it possible to remedy this in any way? It's truly a shame, the ATRAC version of the music sounds so much nicer.

Re: Merging two different encodes to improve quality

Reply #6 –
In general I don't think it may work without some magic estimation of "quality" which doesn't need the original lossless reference for comparison. 

> one which is a mono 24kHz adpcm

if you could be sure that this piece is perfect given the limitations of the format, you could do the following (let's call it A and the other piece B):

1) resample A to 44.1 kHz
2) align them so they match temporally; if it doesn't work out, then tough luck, nothing would realistically work
3) split B to mid / side components
4) find where is the actual high frequency extension of A, let's call it F
5) apply sharp linear phase highpass filter to the mid component of B with cutoff frequency F
6) add A to the mid component of B (if necesasry, change gain of A if it's not right)
7) mix B back to left/right

in other words, you're just removing from B what is contained in A and add it back from A, assuming that it's better.

even then, if the alignment is not perfect, it might just be even worse than the original
a fan of AutoEq + Meier Crossfeed

Re: Merging two different encodes to improve quality

Reply #7 –
Quote from: magicgoose on
In general I don't think it may work without some magic estimation of "quality" which doesn't need the original lossless reference for comparison. 

> one which is a mono 24kHz adpcm

if you could be sure that this piece is perfect given the limitations of the format, you could do the following (let's call it A and the other piece B):

1) resample A to 44.1 kHz
2) align them so they match temporally; if it doesn't work out, then tough luck, nothing would realistically work
3) split B to mid / side components
4) find where is the actual high frequency extension of A, let's call it F
5) apply sharp linear phase highpass filter to the mid component of B with cutoff frequency F
6) add A to the mid component of B (if necesasry, change gain of A if it's not right)
7) mix B back to left/right

in other words, you're just removing from B what is contained in A and add it back from A, assuming that it's better.

even then, if the alignment is not perfect, it might just be even worse than the original
Thanks for the idea. The only flaw I can point out is that the mono version of the music appears to be a downmix of the stereo master, so wouldn't the sound stage get muddled?

Re: Merging two different encodes to improve quality

Reply #8 –
why not transcode both to wave files and then add them as tracks in Audacity (or another similar audio program) in the same project and export both tracks as one file. This assumes that they are exactly matched duration wise though, so no out of syncing would occur.

Re: Merging two different encodes to improve quality

Reply #9 –
Quote from: jaybeee on
why not transcode both to wave files and then add them as tracks in Audacity (or another similar audio program) in the same project and export both tracks as one file. This assumes that they are exactly matched duration wise though, so no out of syncing would occur.
Thank you. I wish it was this simple! This would cause quite an imbalance though, given that one of the files is mono. I have managed to line the music up myself but the only success I've had so far is inverting the difference between the tracks. And even then, the result is funky because the mp3's stereo width is completely compromised by compression. It seems all of the extra information I'm adding is also proportionally adding to the artifacts.

Re: Merging two different encodes to improve quality

Reply #10 –
Quote from: marinawoo on
Quote from: magicgoose on
In general I don't think it may work without some magic estimation of "quality" which doesn't need the original lossless reference for comparison. 

> one which is a mono 24kHz adpcm

if you could be sure that this piece is perfect given the limitations of the format, you could do the following (let's call it A and the other piece B):

1) resample A to 44.1 kHz
2) align them so they match temporally; if it doesn't work out, then tough luck, nothing would realistically work
3) split B to mid / side components
4) find where is the actual high frequency extension of A, let's call it F
5) apply sharp linear phase highpass filter to the mid component of B with cutoff frequency F
6) add A to the mid component of B (if necesasry, change gain of A if it's not right)
7) mix B back to left/right

in other words, you're just removing from B what is contained in A and add it back from A, assuming that it's better.

even then, if the alignment is not perfect, it might just be even worse than the original
Thanks for the idea. The only flaw I can point out is that the mono version of the music appears to be a downmix of the stereo master, so wouldn't the sound stage get muddled?
The mono file is also a mid channel, as in mid/side the mid is the sum of both left and right channels, that's why you do this operation with mid/side. The stereo image comes from the side component, it remains untouched following magicgoose's instructions.

You'll probably need a daw with a plugin for LR to MS mix (ex.: fl studio's stereo shaper).

Re: Merging two different encodes to improve quality

Reply #11 –
Quote from: MetaPixel on
The mono file is also a mid channel, as in mid/side the mid is the sum of both left and right channels, that's why you do this operation with mid/side. The stereo image comes from the side component, it remains untouched following magicgoose's instructions.

You'll probably need a daw with a plugin for LR to MS mix (ex.: fl studio's stereo shaper).
Oh my gosh. Thank you both!! I was struggling to wrap my head around how all this channel balancing works, but I think I understand it now. I followed magicgoose's instructions and the music sounds SO much better. I even managed to change the speed of the ATRAC versions so that all of the music matches up perfectly. Balancing both the ATRAC and ADPCM versions reduced a lot of the ADPCM's artifacts, too. Thank you all so much for your help <3