MP3/AT3 Voicebank Compression [Deleted]

Status
Not open for further replies.

Damien

Wannabe Worker
Retired User
Supporter
Defender of Defoko
Damien submitted a new resource:

Lame/AT3 Voicebank Compression - Compress your voicebanks to unbelievable sizes, uhuhu

Me and Jeffrey have "created" a new way to compress your voicebanks! Now when I say we created, I really mean we're using already existing compression methods, however they haven't been put to good use in the UTAU community until now! This will be a briefing on what exactly we have made and tutorial on how to do this yourself.

We currently have two methods for compression, one of which you can use, and the other we cannot...

Read more about this resource...
 

Jeffrey

local hiatus'er
Defender of Defoko
slight edit: If you want to delete the LAME files badum tss, you can add these lines.

Code:
del /q lame_enc.dll
del /q lame.exe
del /q "decode mp3.bat" >nul 2>nul
rem this is where the batch deletes itself. hopefully no error messages will pop up.
i wanna know how to put progress bar, but forfiles won't let me
 

dyne 智優

Shoutbox Lurker™
Defender of Defoko
in the year 3000, where utau voicebanks that were 300 terabytes are now 10 bytes
in all seriousness, ill remember to use this when i make an utau! which is possibly never but who knows HA
 
  • Like
Reactions: Jeffrey and Damien

Melomad

Ritsu's Renegades
Defender of Defoko
I'm sending to yall boxes of love manifactured in France, it's the best quality love you can get.
More seriously, good job it's sooo great !
 
  • Like
Reactions: Jeffrey and Damien

na4a4a

Outwardly Opinionated and Harshly Critical
Supporter
Defender of Defoko
This will lower the quality of the voice banks unnecessarily.
Your compression recommendation of 10 is a VERY low bitrate that will completely ruin your samples.
If you want to retain the full frequency range of the voice you should use something like FLAC which will give you 50% compression or better (usually better) and won't kill everything.

Also this adds extra steps to extracting the voicebank that no one really wants to do.
 

Damien

Wannabe Worker
Retired User
Supporter
Defender of Defoko
Thread starter
you are literally destroying the quality of your voicebanks
I can see why you say that...
All I can really say against this is that, I didn't hear a difference whenever doing my own experimentation. The initial uncompressed voicebank had the same quality as the compressed and then decompressed voicebank.
 

na4a4a

Outwardly Opinionated and Harshly Critical
Supporter
Defender of Defoko
I can see why you say that...
All I can really say against this is that, I didn't hear a difference whenever doing my own experimentation. The initial uncompressed voicebank had the same quality as the compressed and then decompressed voicebank.
If you don't notice a difference then that generally means that the initial samples are already low quality to begin with. (or your samples have a lot of silence)

There are better compression techniques that get you even better compression without destroying everything.
 
  • Like
Reactions: Kiyoteru and Zoku

Jeffrey

local hiatus'er
Defender of Defoko
I admit, caustic, we didn't really test it on high quality voicebanks, nor we didn't really know the borders of high quality. The only thing on our minds are automated compression of voicebanks.
Also this adds extra steps to extracting the voicebank that no one really wants to do.
I would want you to explain this, we need input that can improve our works.
 
  • Like
Reactions: dyne 智優

Zoku

making doper vocaloid music than the rest
Defender of Defoko
IIRC mp3 has been said by its own inventors to be (currently) the most common worst way to compress audio files even at its highest settings because the format is built on old knowledge and tech. they pulled support on it and while theyre ok with mp3 being used, theyd rather people use things like flac.

so if mp3's own inventors dont really want ppl to use mp3 anymore because its lower quality inherently, why should you do it on voicebanks where people seek high quality samples in voicebanks
 

na4a4a

Outwardly Opinionated and Harshly Critical
Supporter
Defender of Defoko
If you are going to add extra steps then it needs to be auotmated, most people just want to drag and drop the voicebank and be done with it.

Also a 100-200MB voicebank is nothing. Worry about compression when your voicebank reaches 700-900MB.
 
  • Like
Reactions: Kiyoteru and Zoku

Jeffrey

local hiatus'er
Defender of Defoko
okay, non-mp3 file that can compress in a lossless way, and automated conversion ala-moresampler style, coming right up soon.
definition of soon: lacking experience, kind of hard, maybe next month/year
we do still have the rights to keep this thing up, right?
@_caustic_
This will lower the quality of the voice banks unnecessarily.
Your compression recommendation of 10 is a VERY low bitrate that will completely ruin your samples.
If you want to retain the full frequency range of the voice you should use something like FLAC which will give you 50% compression or better (usually better) and won't kill everything.

also, i did some reversion on the lame.exe quality control. The original is:
0 = highest quality
9 = lowest quality
x=10-(input)
so 10 is actually 0
 
Last edited:

Zoku

making doper vocaloid music than the rest
Defender of Defoko
okay, non-mp3 file that can compress in a lossless way, and automated conversion ala-moresampler style, coming right up soon.
definition of soon: lacking experience, kind of hard, maybe next month/year
js compression is neat but you dont exactly need it in (relatively) small vbs, which most of us have.
 
  • Like
Reactions: Kiyoteru

Zoku

making doper vocaloid music than the rest
Defender of Defoko
'js' is short hand for 'just saying'

i was 'just saying' that compression is neat
 

na4a4a

Outwardly Opinionated and Harshly Critical
Supporter
Defender of Defoko
well, compression can help a lot like the place where i live, where internet is effing slow you can't bear to live
It's better to wait and get what you expect then to get something fast and inferior.

Also with FLAC compression you should be fine.
Or even simpler, if you use 7zip it probably will already get you there without extra stuff to do.

Most voicebanks are big because people fail to delete star, platinum, spectrum, etc, files. Not just because of the actual audio content.
 
Status
Not open for further replies.