What are the most complete and intuitive Reclists that you would recommend ?

Pikachuk(SeleDreamsP)

Teto is love, Teto is life (Rin too)
Defender of Defoko
74
Hi, I'm Pikachuk, I'm currently working on the voicebank creation and management software and this is why I'd need some suggestions from the community.

I am currently searching the most complete and intuitive reclists that could be used as the standard system for the software.
The most important ones I'm searching for are at the moment :
  • Japanese CV
  • Japanese VCV
  • English CV-VC
  • English VC-CV
  • English ARPASING
But I think it would be interesting to add more languages too such as french and spanish.

The important actually for the reclists is to be at the same time complete to cover most cases while being intuitive to use for the end users.
 

Kiyoteru

Local Sensei
Supporter
Defender of Defoko
The precise specifications of a a reclist depends on the individual user, and the freedom that UTAU currently has is what I believe to be one of its greatest strength in comparison to other singing synthesis software. To decide on the "best reclist for everyone" is not a question that can be answered in one post, but necessitates a survey of or discussion between the entirety of the community.

That said, "VCCV English" is the name for only one reclist, which is PaintedCZ's 2015 English reclist. If you seek to eliminate the numerous variations and customizations that people have made for their own work, look no further than the original author and ensure that you keep up with information regarding their planned update.

In terms of Arpasing, you can choose to use the official computer-generated reclist, but this is the reclist that most Arpasing opponents use to define all of the method's faults. Once again, customization by individual users is the factor that brings out the strength of the method.

For English CVVC, this is a broad term that can refer to virtually every English reclist. In the interest of standardization, I recommend Delta English version 5. This will, however, ignore all independent reclist writers. Their work will not be able to reach a potential audience that may find a niche reclist to perfectly suit their needs.

Japanese reclists have less variation than English due to the nature of the language's phonology and notation, but there is still the question of what extra samples to include. A reclist that covers every essential Japanese sound is complete, yes, but what about stylistic variations in pronunciation, such as "ng" samples that replace "g"s or rolled Rs in addition standard taps? What about samples for non-Japanese words, are those essential to
what is supposed to be a Japanese-only voicebank? And common expressive samples, such as breaths and fries? For many beginning users, these samples can easily be ignored, but for advanced users, these samples are essential to their musical and stylistic goals and a voicebank wouldn't reach full potential without them.

intuitive to use for the end users.

What type of end users are we targeting? The largest factor in reclist design decisions is whether to suit voicebank developers or voicebank users. This can take a reclist to extreme opposite ends of a spectrum. A reclist meant for voicebank developers may try to reduce the number of samples recorded and the amount of configuration needed, in order to make voicebank production faster and easier to complete. But this will come at the sacrifice of usability, where, without automatic tools, will make editing USTs a very long and arduous task (for example, Japanese CVVC or Arpasing). A reclist meant for voicebank users will likely require the voicer to record and configure many more complex samples, taking up more precious disk space or OTO space and reducing the number of pitches that can be included in a single voicebank. However, it can reduce the amount of work when editing USTs, allowing the user to use fewer, longer notes, and focus more on musical tuning decisions.

In the end, I guess the reclist that wins will just be the most popular ones, and there will always be users who are frustrated with the features that are sacrificed in the name of other goals. You risk shutting out innovation.
 

Pikachuk(SeleDreamsP)

Teto is love, Teto is life (Rin too)
Defender of Defoko
Thread starter
74
The precise specifications of a a reclist depends on the individual user, and the freedom that UTAU currently has is what I believe to be one of its greatest strength in comparison to other singing synthesis software. To decide on the "best reclist for everyone" is not a question that can be answered in one post, but necessitates a survey of or discussion between the entirety of the community.

That said, "VCCV English" is the name for only one reclist, which is PaintedCZ's 2015 English reclist. If you seek to eliminate the numerous variations and customizations that people have made for their own work, look no further than the original author and ensure that you keep up with information regarding their planned update.

In terms of Arpasing, you can choose to use the official computer-generated reclist, but this is the reclist that most Arpasing opponents use to define all of the method's faults. Once again, customization by individual users is the factor that brings out the strength of the method.

For English CVVC, this is a broad term that can refer to virtually every English reclist. In the interest of standardization, I recommend Delta English version 5. This will, however, ignore all independent reclist writers. Their work will not be able to reach a potential audience that may find a niche reclist to perfectly suit their needs.

Japanese reclists have less variation than English due to the nature of the language's phonology and notation, but there is still the question of what extra samples to include. A reclist that covers every essential Japanese sound is complete, yes, but what about stylistic variations in pronunciation, such as "ng" samples that replace "g"s or rolled Rs in addition standard taps? What about samples for non-Japanese words, are those essential to
what is supposed to be a Japanese-only voicebank? And common expressive samples, such as breaths and fries? For many beginning users, these samples can easily be ignored, but for advanced users, these samples are essential to their musical and stylistic goals and a voicebank wouldn't reach full potential without them.



What type of end users are we targeting? The largest factor in reclist design decisions is whether to suit voicebank developers or voicebank users. This can take a reclist to extreme opposite ends of a spectrum. A reclist meant for voicebank developers may try to reduce the number of samples recorded and the amount of configuration needed, in order to make voicebank production faster and easier to complete. But this will come at the sacrifice of usability, where, without automatic tools, will make editing USTs a very long and arduous task (for example, Japanese CVVC or Arpasing). A reclist meant for voicebank users will likely require the voicer to record and configure many more complex samples, taking up more precious disk space or OTO space and reducing the number of pitches that can be included in a single voicebank. However, it can reduce the amount of work when editing USTs, allowing the user to use fewer, longer notes, and focus more on musical tuning decisions.

In the end, I guess the reclist that wins will just be the most popular ones, and there will always be users who are frustrated with the features that are sacrificed in the name of other goals. You risk shutting out innovation.
I probably didn't word it well

When I talked about the most intuitive and complete, it wasn't to say *the* most but mostly suggestions of reclists able to cover most of the cases required by the language while not requiring a syntax too complex to learn

My software is planned to make the UTAU voicebank creation more assisted and standardized in order to make it easier for UTAU newcommers and the end users

about the "facultative" samples, I talked about it on the thread of my project, the software will detect the facultative samples and will allow to chose between the "full" or "core" mode to chose if we want to record all the samples or only the essential ones
 

Pikachuk(SeleDreamsP)

Teto is love, Teto is life (Rin too)
Defender of Defoko
Thread starter
74
I guess I'll have to chose by myself the reclists, I thought adding some debate to the reclists would have helped taking the best decisions for the software
 

Pikachuk(SeleDreamsP)

Teto is love, Teto is life (Rin too)
Defender of Defoko
Thread starter
74
It is still too soon to be making a decision. You should branch out to other communities to gather more data. I'll share this topic on Discord to encourage further discussion.
I see
and it's because mostly, I needed at least one reclist to continue development since I was working on the project system that determines the OTO etc from the reclist
that's why I asked this pretty soon
 

Kiyoteru

Local Sensei
Supporter
Defender of Defoko
In the case of VCCV English, there is only one VCCV English reclist, which you can use now for software testing purposes.

As for other languages and methods, it will require more deliberation to decide what reclist should be set as the definitive standard for a majority of voicebanks going forward.