This thread is to show Dynamivox's DiRS project.
DiRS is a project to create recording lists that are universal between speakers within a language without breaking compatibility between voicebanks from different regions.
Designed to be modular so users can swap between languages and voicebanks without having to relearn a phonetic system. DiRS is based on X-Sampa with some simplification and Utau-compatible changes.
To make it as easy as possible to record, we want to get examples from as many users from around the world to build up a demonstration catalog. Users wishing to record the list can use the catalog as reference so they know how each phoneme roughly aligns to their way of speaking.
No need to edit the reclist here!
DiRS is meant to be standardized to avoid confusion and new lists have to be created, reviewed, and approved by Dynamivox. This is to avoid confusion between users from unofficial alterations. We want to work with native speakers if possible in order to correctly understand the language structure.
When changes are made to the lists after release an upgrade path will be available so voicebanks can be updated to the latest standard with minimal effort.
The first language for DiRS to be developed is English. The list is made with non-Americans in mind and how pronunciation shifts when singing. As such, some vowels may be missing or different compared to other lists but creates a much better result.
Speakers from the United States can still use this list without any problems. An EN-US addon set may be released with the base list if deemed necessary.
Here are some demos of DiRS-en in action!
ITALIAN VOICER DEMO (makku):
UK VOICER DEMO: (coming soon)
SPANISH VOICER DEMO: (coming soon)
US VOICER DEMO: (coming soon)
(others coming soon)
DiRS allows for what are called "implicit" and "explicit" compounding.
An "explicit" compound is the combination of similar languages that are closely related to not risk conflicts. This is usually limited to languages with similar origins. These can often be built by users without help from Dynamivox.
An "implicit" compound is more complicated, it's the combinations of languages that are not normally compatible but can be merged do to a user's regional way of speaking. These will require Dynamivox to intervene to build the list, if deemed possible.
---
WE ARE LOOKING FOR DEMO VOICERS!
Since DiRS is made for users around the world, we need testers from as many locations as possible! As of right now we are not interested in users from the United States.
You will be given a fragment of the list to record, at this time we want to avoid having the full list recorded in case large changes are made.
EDIT: Follow the instructions on my post here: http://utaforum.net/posts/125118/
---
Q&A:
(none here right now)
Feel free to ask questions! I'll be glad to answer them!
Disclaimer: DiRS is not an alternative to not knowing a language (i this case English), if you have absolutely no understanding of how a language is pronounced then no reclist will work for you. While the future pronunciation catalog will help, you will need some understanding.
DiRS is a project to create recording lists that are universal between speakers within a language without breaking compatibility between voicebanks from different regions.
Designed to be modular so users can swap between languages and voicebanks without having to relearn a phonetic system. DiRS is based on X-Sampa with some simplification and Utau-compatible changes.
To make it as easy as possible to record, we want to get examples from as many users from around the world to build up a demonstration catalog. Users wishing to record the list can use the catalog as reference so they know how each phoneme roughly aligns to their way of speaking.
No need to edit the reclist here!
DiRS is meant to be standardized to avoid confusion and new lists have to be created, reviewed, and approved by Dynamivox. This is to avoid confusion between users from unofficial alterations. We want to work with native speakers if possible in order to correctly understand the language structure.
When changes are made to the lists after release an upgrade path will be available so voicebanks can be updated to the latest standard with minimal effort.
The first language for DiRS to be developed is English. The list is made with non-Americans in mind and how pronunciation shifts when singing. As such, some vowels may be missing or different compared to other lists but creates a much better result.
Speakers from the United States can still use this list without any problems. An EN-US addon set may be released with the base list if deemed necessary.
Here are some demos of DiRS-en in action!
ITALIAN VOICER DEMO (makku):
UK VOICER DEMO: (coming soon)
SPANISH VOICER DEMO: (coming soon)
US VOICER DEMO: (coming soon)
(others coming soon)
An Arpasing branch of DiRS is in the works but it will not be the main version. the Arpasing conversion requires a lot of feature dropping to be compatible and should be used as an alternative reclist for those wanting to use the Arpasing tools and method instead. We can prepare a kit at a later date that lets users convert standard DiRS-en (xsampa banks) to DiRS-en-Arpasing at the expense of stripping a large portion of the functionality.
The Arpasing version will not work with the Arpasing Assistant and will require a special pseudo-bank to generate the phonemes. At this point you can load your voicebank and edit from there.
This may change if the Arpasing tools are updated for more customization.
The Arpasing version will not work with the Arpasing Assistant and will require a special pseudo-bank to generate the phonemes. At this point you can load your voicebank and edit from there.
This may change if the Arpasing tools are updated for more customization.
DiRS allows for what are called "implicit" and "explicit" compounding.
An "explicit" compound is the combination of similar languages that are closely related to not risk conflicts. This is usually limited to languages with similar origins. These can often be built by users without help from Dynamivox.
An "implicit" compound is more complicated, it's the combinations of languages that are not normally compatible but can be merged do to a user's regional way of speaking. These will require Dynamivox to intervene to build the list, if deemed possible.
---
WE ARE LOOKING FOR DEMO VOICERS!
Since DiRS is made for users around the world, we need testers from as many locations as possible! As of right now we are not interested in users from the United States.
You will be given a fragment of the list to record, at this time we want to avoid having the full list recorded in case large changes are made.
EDIT: Follow the instructions on my post here: http://utaforum.net/posts/125118/
---
Q&A:
(none here right now)
Feel free to ask questions! I'll be glad to answer them!
Disclaimer: DiRS is not an alternative to not knowing a language (i this case English), if you have absolutely no understanding of how a language is pronounced then no reclist will work for you. While the future pronunciation catalog will help, you will need some understanding.
Last edited: