UTAU "Resampler" Directory

shinami

Procrastination Queen
Tutor
Supporter
Defender of Defoko
This thread needs help! Please read this if you want to know what help I need.
This thread still needs some work. Anyone who has knowledge on EFB-GW or favoco6 - PLEASE share it. Also, if I missed any flag differences for world4utau, please tell me! See my post on the top of Page 4 of this thread for issues I'm currently facing and information I need.

I wanted to make this article to help people confused about the various "resamplers", so please help me polish it up!

(On the matter of how I reuploaded the old version of TIPS, I'm currently working on contacting ScientistB to gain permission to do so/continue doing so.)

Quick Intro - What are "resamplers"?
"Resamplers" are what make UTAU sing, essentially. They, with the help of wavtool, take your samples and what's on the UST and make the samples sing the notes. Resamplers also generate frq files (pmk files for TIPS and frc files for M4) that allow UTAU to quickly go through the process of rendering to sing. When using UTAU, you can tell what resampler you are using and easily change it here:

2qu7wph.png


If you ever open a UST and what's listed there is red, just change it to a resampler you have or change the directory to where that resampler is located.

Why do we call them all resamplers? It's because that area - Tool 2 - is where you put the .exe that will "resample" the wav samples, thus creating the name "resampler". Resampler was the very first resampler for UTAU, and though since then there have been many more invented that are not called Resampler, we still catagorize them as a whole as "resamplers".

Each resampler is different from the others and each has voicebanks it's good with and voicebanks it's bad with. Each has specific behaviors, different flag setups, different file creation, and more. So just download some, put them into UTAU, set your UST to use that resampler, and experiment! Every UTAU can sound very different with the various resamplers, so experimenting is key to finding the most compatible resampler for your bank.



You can get all publicly available "resamplers" here:
http://www20.atwiki.jp/utaou/pages/36.html#id_41c7e81e

For those who don't understand the chart/are too lazy to go on a Japanese site, here's a list of them, their download links, and any special notes on them.

Released "Resamplers"

resampler: Latest version always comes with the latest version of UTAU. Default "resampler". Current version is resampler with 4.12 while the still good resampler10 is with 2.79/2.77 and resampler11 was beta'd with 2.90.
(Download UTAU 2.76 ZIP here: http://utau2008.xrea.jp/utau0276.zip
You can upgrade to UTAU 4.12 here (No ZIP available): http://utau2008.xrea.jp/utau0412installer.zip)

Notes: Standard, basic resampler everyone knows and has. Follows the standard flag uses. HOWEVER as of the latest resampler in 4.12 B/BRE is handled differently; B0 creates a similar effect to TIPS while very high B values can make all voices breathy and whispery, even strong banks.

fresamp: Current version is 0.11 - http://bit.ly/ae63qb
Notes: Is known for eliminating noise, strengthening voices, and its nasal effects on higher notes. Also known for taking a very long time to render in comparison to other resamplers. Flag differences are that F controls the format filter (F1 or F0 are highly recommended when using fresamp) and L controls turning the format filter on and off. Please see this chart for what flags are usable with fresamp: http://www10.atwiki.jp/utau2008/pages/39.html

TIPS: Current version is Beta - http://scientistb.web.fc2.com/program/ (Click "TIPS")
Notes: Is known for being good with softer banks and low notes, but can turn banks metallic and can overly emphasize noise. Creates pmk files instead of the usual frq files. Flag differences are that only H, P, t, g, and R work with TIPS currently; R forces samples with this flag to regenerate their pmk files.

TIPS (older version): No longer available on ScientistB's page because TIPS has been upgraded since this edition. However, I feel they're very different and older TIPS has some advantages over newer TIPS, so I reuploaded it. - Link taken down until permission from ScientistB is gained
Notes: Is known for working well with low notes and strengthening soft voices alongside being great with male vocals, but also can emphasize noise. Does not accept the pmk files generated by newer TIPS, so you have to delete and regenerate them. Flag differences are that it doesn't accept any flags at all.

EFB-GW: Also known as WORLD. Current version is 0.1.2 - http://www.slp.is.ritsumei.ac.jp/~morise/world/world-0.1.2.zip
Requires FFTW files to work. Download here - http://www.fftw.org/install/windows.html
Notes: I am unknowledgable about EFB-GW since I cannot work it. Information will be provided when I recieve input from those who can. All I currently know is that EFB-GW apparently doesn't work on computers outside of Japan at the moment.

world4utau: Also known as w4u. Current version is 0.0.1 - http://utau2008.xrea.jp/mp3/engine_hikaku.html
Requires FFTW files to work. Download here - http://www.fftw.org/install/windows.html
Notes: Is known for being very, very picky regarding which banks work with it (More likely to work with mono samples recorded at 41000Hz 16bit with no metadata). With banks it does work with, it can create odd noise if there's the slightest noise in samples, but banks without noise come out stronger and clear. I am not knowledgable about any flag differences.

bkh01: Current version is 0.55 - http://z-server.game.coocan.jp/utau/bkh01_055.zip
Notes: Is known for creating more realistic voices with the right flags and voicebanks and working well with softer voices, but also can create distortion, metallic and nasal voices, and odd cracking on stronger voices and other voices not suitable for it. A certain level of breathiness is needed in a bank to sound good with bkh01. Flag differences are that is uses the A, S, and V flags. Combining a low b value and a higher V value is recommended when using bkh01.

tn_fnds: Current version is 0.0.8 - http://z-server.game.coocan.jp/utau/tn_fnds008.zip
Requires FFTW files to work. Download here - http://www.fftw.org/install/windows.html
Notes: Is known for functioning similarly to old TIPS in that it works best on low notes and strengthens, but unlike TIPS it doesn't magnify noise. However, currently there are issues in that it makes many banks jumpy or off pitch on some notes. Also does not accept stereo samples. Flag differences are that it only accepts g, A, t, B, O, and e (e changes its sample method from looping to stretching like most resamplers - may be very useful in preventing jumpiness).

EFB-PB: Current version is from 7/21/12 - http://www1.axfc.net/uploader/so/2537472
Notes: Requires EFB-GW to work. Likely works well with soft voices. I need to update this.

EFB-GT: Current version is from 11/22/12 - http://www1.axfc.net/uploader/so/2687150
Notes: Works very well with soft voices, smoothing out noise while making consonants crisp. Gets a bit weak in the higher ranges but that may vary UTAU to UTAU. Still need to update this more.

Unreleased "Resamplers"

M4: Also known as model4. Currently in closed beta. Ameya has stated that it will never be released because it is too unstable.
Notes: I do not have M4, so this is all from current public knowledge on the beta.
Creates frc files instead of frq or pmk. Is known for a few glitches, including how sounds without a coherent pitch cannot render an frc, most notably being breath samples. Also known for being more realstically breathy than resampler which can be managed with flags and also managing pitchbends better. Generally makes voices soft and breathier. It has mostly the same flags as resampler, but H, h, C, D, E, and b don't work by current knowledge.

favoco6: Also known as phavoco. Currently in very closed beta.
Notes: I do not have favoco6, nor have I heard anything about it, so I am unknowledgable about it. Information will be provided when I recieve input from those who are.
 

Iscabird

Ritsu's Renegades
Defender of Defoko
Using B does work with fresamp. Sort of. All it manages to do with my UTAUs is cause some notes to buzz horridly (particularly problematic with VCVs as a different note aliased from the same sample may not buzz at all) or just go silent. b works as well but personally, I don't use it.
 

Cdra

possibly dead
Global Mod
Supporter
Defender of Defoko
world4utau... is a weird, finicky thing.  It doesn't render for all banks.  It sounds bad with a lot of banks, causing weird noises in the recordings; however, banks recorded on higher-quality microphones with very little background noise tend to work very well with it.  It generates clear, smooth sound in these cases; however, it reacts very poorly to any noise in a sample.  It can also make soft banks sound a bit stronger.

I've experimented with it some, but I'm going to keep doing so.  This is just my observations on it so far.
 

kirrinx

Teto's Territory
One of my older friends has UTAU and has the model4 Resampler and from what she has told me and from what I've heard of the samples , it wont necessarily work for every VB but it does help Male and Female VB's sing lower notes much better just as it would with TIPS ~


for the flags that do work with it are g , Y, H , B , i haven't confirmed any others ~

Help makes soft voices sound smoother and Stronger voices get a bit weakened when they are rendered with this resampler ~
 

shinami

Procrastination Queen
Tutor
Supporter
Defender of Defoko
Thread starter
Ah, thank you all!

Clearly I was mistaken about some things regarding M4. It's been a really long time since I read the thread for it back on the old forum. Do you know any flags that have different functions/don't work, Kirrin, or are you saying that those flags are the only ones that work?

Also, thanks Kiidos. I've tried b and B, but they never made a difference so I assumed they didn't work. I'll change that they work minimally compared to resampler.

Ah, and thank you sooo much for all that information Cdra! *v* Like I asked with M4, do you know if any flags function differently/don't work/it's like TIPS and is limited?

Edited all the information so far in~
 

kirrinx

Teto's Territory
No those are the only ones im aware that do work xD with M4 ,

idk if C, L , F , h , b , x , t, E, W  or others work i'll ask her to test them out and ill edit this post * later on ~
 

shinami

Procrastination Queen
Tutor
Supporter
Defender of Defoko
Thread starter
Added the new tn_fnds resampler. I'll have to watch it carefully for updates and changes in its functions.

Also, fortunately, because of getting tn_fnds, I've finally gotten the correct FFTW files to also run world4utau and EFB-GW. Hopefully I'll be able to get some banks to cooperate with them to get some basic obvservations of my own. *u*

Edit: I stand corrected, EFB-GW still refuses to work.
 

tady159

Ruko's Ruffians
I have talked quite a lot about EFB-GW with Masao, and... as it seems, it only works on computers with Japanese environment. We were almost getting it to work in other cases, but when I was going to give him feedback, my computer crashed >o<

As far as I have gathered information, w4u is more likely to work (but I'm not sure if it will) with mono samples recorded with ratings at 41000Hz and 16 bit, and with NO metadata. Plus all those dlls we have to download.
 

shinami

Procrastination Queen
Tutor
Supporter
Defender of Defoko
Thread starter
That's unfortunate. Hopefully it'll get an update soon.

Though with how world4utau and tn_fnds function rather similarly (At least with banks I tested - there are some small differences though) and they both use those .dlls, I almost think EFB-GW sounds similar to them probably.

But thanks, I'll add that information to the EFB-GW section.
 

tady159

Ruko's Ruffians
Well, both tn_fnds and w4u ARE based on EFB-GW, but I believe that w4u sounds closer to it. But I'm not sure.
 

Dangosan

Jellie Bellie Pete Rat Gummie Candie
Defender of Defoko
@Shinami Did you get permission from ScientistB to redistribute TIPS?
And also, UTAU 2.77 is public. It's available at http://utau-synth.com. This page contains the latest Win and Mac versions. Only the 2.90 version is private.
 

shinami

Procrastination Queen
Tutor
Supporter
Defender of Defoko
Thread starter
Dangosan link said:
@Shinami Did you get permission from ScientistB to redistribute TIPS?
And also, UTAU 2.77 is public. It's available at http://utau-synth.com. This page contains the latest Win and Mac versions. Only the 2.90 version is private.

...I haven't. That's why I'm intensly worried about giving it out. I'd love to contact and ask, but I'm afraid of the language barrier and that I'd be unable to express how different I feel older TIPS and newer TIPS are and how I'm trying my best to make sure people are educated as to the differenes. If you think they would 100% dislike me giving it out I'll happily (Well, with a sigh moreso because I really like older TIPS better) take it down.

And I'll add that link. But I know some people prefer to have UTAU ZIP and 2.77 isn't provided in ZIP, so I'll also leave the link to 2.76. Thanks for showing me where it is though! I was wondering how people had 2.77.


As a small note on the thing as a whole to anyone who cares, I'm going to add a section explaining what "resamplers" are for those who don't know and I'll put the asking for help part in a spoiler.
 

tady159

Ruko's Ruffians
M4: Also known as model4. Currently in closed beta alongside UTAU 2.77.

M4 does not come WITH any version of UTAU yet, and the closed beta version is 0.2.9, and not 0.2.77, which is already available for public download.

And as a matter of fact, if you have, for example, 0.2.76 ZIP, and you install 0.2.7, you can also copy the new UTAU version (which is essentially the only difference) to the ZIP folder, and it will work as if you had downloaded 0.2.77 as a ZIP.
 

shinami

Procrastination Queen
Tutor
Supporter
Defender of Defoko
Thread starter
tady159 link said:
M4: Also known as model4. Currently in closed beta alongside UTAU 2.77.

M4 does not come WITH any version of UTAU yet, and the closed beta version is 0.2.9, and not 0.2.77, which is already available for public download.

And as a matter of fact, if you have, for example, 0.2.76 ZIP, and you install 0.2.7, you can also copy the new UTAU version (which is essentially the only difference) to the ZIP folder, and it will work as if you had downloaded 0.2.77 as a ZIP.

...I feel so silly now. I get all my facts mixed up or messed up sometimes. Thanks for clarifying, sorry you had to.

And I figured that out shortly after typing my response to Dangosan. I really just was advising people to downlaod the ZIP because this a resampler directory, not a UTAU program directory, and they can easily take resampler from the ZIP folder (unless it's been updated in 2.77...? I doubt it). But I implied it now with my edit that you can just upgrade to 2.77.


Ahh, as another random update, now adding the changes to tn_fnds.
 

tady159

Ruko's Ruffians
OK, I read the spoilers at the beginning, and I will try to help with this thread a bit.
I will try to go in order:
About EFB-GW, I already gave all the information that I could, since it doesn't work for me at all.
I have no information to give about favoco, sadly.
On bkh01, with B0 and b0 setting, it is able to extend the voices to the whole piano roll, but the results are still not very good, specially because it DEPENDS on a certain amount of breathness on the voice to sound natural, and also, the lower the value on breathness, the more it will damage consonants (specially "t" and "k"). Its breathness settings sound more realistic than the other resamplers, and it is the only engine that gives full control of it. Differently from resampler, only using B100 will not make the voice go "unvoiced" and just whisper (mostly in a creepy way), but you would need to set V to 0 too to get this result.

As of M4, it has mostly the same flags as resampler, except H, h, C, D, E and b (those ones were all that I could find) also, it is hard to know which flags it accepts, cause it doesn't have an official list. Its formant filter can't be turned off. As of handling voices, it is still a bit unpredictable, but in a general rule, it will make voices a bit softer and beathy. It creates its own frequency files (extension .frc), but it has a bug which it doesn't read frequency of breaths samples, and crashes (if the breath is long enough, it may read correctly though). It handles pitchbends better than default resampler.

According to TIPS' readme file, the flags it handles are g, H, P and t (same usage as default resampler), and a unique flag, R, which forces TIPS to generate .pmk files.
About redistributing the old TIPS, I would not tell you to take it down, but this is a quote from TIPS' readme: "If you want to redistribute TIPS.exe, prease ask me for the permission.
  The reason is that if I permit unlimited redistribution,
  old version program with defect may be left somewhere." (this English is not mine, but the creator's him/herself)

The current resampler version is 10. I don't know since WHICH version it is being distributed though. resampler11 is available as closed beta, and THIS ONE, comes included with UTAU version 0.2.9.

As of fresamp flags, there is this page which has a comparison table for flags available for old versions of resampler, and fresamp http://www10.atwiki.jp/utau2008/pages/39.html

(note: You left the current version of tn_fnds as 0.0.2)
 

shinami

Procrastination Queen
Tutor
Supporter
Defender of Defoko
Thread starter
@tady Oh gosh, thank you so much! ; o ; So many little inconsistencies I need ot fix, and then some new information...! Seriously, I cannot thank you enough for putting the time in to type all that out and look around at all those files. I'll go adding th new information and making fixes now.

And that a relief that I can ask to redistribute it and it seems they have a strong grasp on English. I can't believe I didn't think to look at TIPS's readme for this information. I'll go asking ifthis is alright right away, and hpefully my English will be clear enough. I don't want to butcher it with Google Translate.

Edit: Wait...fudge...they say to comment on their blog, but there's no way to do that. And I don't have a Twitter. Would this be appropriate to ask over Twitter? I don't have one and I don't know if I'd be able to make a good case for myself with the character limit there.
 

tady159

Ruko's Ruffians
I don't know if I'd be able to make a good case for myself with the character limit there.

I will not try to push you to make a Twitter account, but I can say that you get used with the character limit XDD
 

Dangosan

Jellie Bellie Pete Rat Gummie Candie
Defender of Defoko
@Shinami You need prior permisson before redistributing TIPS.
 

shinami

Procrastination Queen
Tutor
Supporter
Defender of Defoko
Thread starter
Dangosan link said:
@Shinami You need prior permisson before redistributing TIPS.

Yeah, I know. I'm contemplating taking the link down until I get confirmation. I probably will.

But today I'm going to try and set up a Twitter account, figure it out, and then try to figure out how to go about asking for permission to distribute while explaining my ideas in that stupid character limit. Ugh. I really hope they don't mind me asking this over Twitter. *is nervous*
 

Zoku

making doper vocaloid music than the rest
Defender of Defoko
エターナル・フォース・ブリサンプラー ジェントリー・ウィープス
(EFB-GW)
Eternal Force Block Resampler - Gentry Weeping

(Possibly referring to the fact it works well with soft voices. (Weeping = crying = (sometimes) soft)