MP3/AT3 Voicebank Compression [Deleted]

Status
Not open for further replies.

Jeffrey

local hiatus'er
Defender of Defoko
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.

3rd world countries and data caps are factors that encourage people to compress files. If you look inside the batch scripts, you can see that it deletes unnecessary files (like llsm and uspec) after it executes the main function.

7zip is one of the best solutions for creating SFX archives, yes. But we chose WinRAR for its availability at the moment. Experienced users are always welcome to wander from our path. They are also welcome to alter/modify the script to suit their tastes (albeit it's a dangerous thing to do)

We provide these scripts "as-is" and will not be held responsible for damage of the system of the end user. The scripts we provide are safe, and are harmless to the system.

You can also take a look at the FLAC version I recently made with your suggestions (moresampler-esque drag&drop). It's in here: http://utaforum.net/threads/flac-lossless-voicebank-compression.16172/
If you have a better alternative to FLAC, please let me know.

edit: @Zoku, please look at the script and take a look at the input variable. I've been reading the "help" section of the executable, and I am aware that 0 is the highest quality. In the script, 1 is the lowest quality. I did math functions to make it that way. A pretty simple one at that.
 
Last edited:

na4a4a

Outwardly Opinionated and Harshly Critical
Supporter
Defender of Defoko
I've been doing this for years but in a better and prettier way (because people value prettiness over function). But I won't describe that here since you have your own thing going for you.

First is that if you are in a third world country then that's valid but the number of users in those areas is a lot smaller than you may be suggesting.
Second is that you should be using 7zip, it's more available and it's free to use. Where no one buys WinRAR and it can promote piracy.
Third is that many antiviral programs and even Windows itself (via smartscreen) can block unsigned scripts and software you download from the internet.
Forth is that I suggest the use of FFMPEG wherever possible since it's a capable of many formats, would be less to download, and lets you more or less retrofit the same script to any format. It's also fairly easy to understand even without the manual.
Fifth is to make sure you keep frq, frc, mrq, pmk, dio, and vs4ufrq files as they are not very large and avoid having to regenerate or destroy fixed frequency maps.
Sixth is to delete *.star, *.platinum, oto.setParam-Scache, and setParam-backup.ini. the star and platinum files also use a ton of space and can double the size of a voicebank.
Seventh is to use M4A (AAC/MP4 Audio) instead of MP3 for the lossy compressor as it's known for being much better. You can limit the bitrate to 160kbps for good quality and a smaller size. "-b:a 160k"
 

Jeffrey

local hiatus'er
Defender of Defoko
okay this debate is getting longer.
1.) We made this project to favour the minority mentioned.
2.) Again, we are not forcing people to use WinRAR.
3.) Thank you for that note.
4.) FFMPEG can reach up to ~30MB. We don't want to bundle a ~30MB file in our 30MB(post compression, 50%) VB.
5.) Our program does not delete those. We are aware of its importance. We only delete .llsm and .uspec files at the moment.
6.) I did not encounter those files while creating voicebanks. Although that can easily be appended in the batch scripts.
7.) I made a FLAC compressor, so we won't drone over this MP3 compressor. I can present evidence to defend my case, but I reckon it'll make things longer. Thank you for that suggestion.

you might also want to look at the license of m4a/aac encoding/decoding.
 
Last edited:

na4a4a

Outwardly Opinionated and Harshly Critical
Supporter
Defender of Defoko
MP3 was designed for comercial use as well so you should check the licensing for that.
But for small, non-commercial, hobbyist use, you won't have issues. All you are doing is making a command line/bat file to run another program.
 
  • Like
Reactions: Kiyoteru

Jeffrey

local hiatus'er
Defender of Defoko
I'm planning to use a more open-source audio compression format. I just knew about these kind of stuff this weekend.
And yes all I'm doing is making batch scripts to convert necessary audio samples to a lossy file. We did not, after all, claim that we made new lossy compression ourselves.

I hope your initial concern about the post-compression quality is over. I wrote a script that uses FLAC, so people who doesn't like lossy formats are welcome to use/alter/modify/distribute my scripts. People would most likely consider this as another option, not a mandatory step to distribute banks.

I tested the LAME encoder on 若菜マユーDark Energyー and made this:

I'm pretty sure the instrumentals will cover up the difference. EQ, reverb, and the instrumentals are likely to conceal the missing frequencies anyway.
 

na4a4a

Outwardly Opinionated and Harshly Critical
Supporter
Defender of Defoko
testing the FLAC encoder is not necessary, it's been tested by thousands of people and is a true lossless codec. no audio data is lost at any point in time.
The difference between the two is most likely due to dithering which is noise shaping to get rid of aliasing caused by limited bitdepth (16 bit samples).

Also for lossy just use Ogg Vorbis which is highly compatible and light. It's also free to use. There is no need to be all mysterious about what codec you are going to use next.

I never claimed you claimed to make a new format yourself, but your argument about licensing is a moot point since you are not implementing the codec as part of a product.
 

Damien

Wannabe Worker
Retired User
Supporter
Defender of Defoko
Thread starter
This resource has been removed and is no longer available for download.
 
Status
Not open for further replies.