Pathos
patreon
NyVox Update
🕑 Added 2022-11-04 02:55:23 +0000 UTCThis is not the final version, this is mostly just an update to address some fixes and test some things out. It's been a while since I posted an update and I want to keep everyone informed even if things are not done yet. In fact, this updated version may be worse in some ways, so do give it a test first.
The download is now an installer, here is the link: https://nyvoxdata.com/v1/NyVoxSetup.msi
(Please try out the free voices before subscribing to ensure NyVox works with your computer.)
NyVox requires Windows 10 version 1903 or later and a modern GPU (GTX 950 or later) (As a reminder, Steam keys are no longer being distributed but a future Steam release is still planned.)
Changes:
- Fixed many issues with stuttering/skipping. There were a few causes for this. NyVox will attempt to get Realtime priority, which you may need to run as Admin for OR set it manually in task manager while it's converting. This should prevent skipping even with a lot of CPU usage but Setting this is not required and only suggested if you run into issues with CPU usage.
- Added models. Swap between 3 different models, A1, B1, A2. They have different qualities and resource usage. B1 may sound good with some voices, but not with others. Some voices with it have a robotic/metallic sound, so test them before using them. A1 is faster with decent quality. A2 is slightly better quality but has higher "wobble" to it and higher CPU usage.
- Pitch Detection. It's worse in some ways, unfortunately. It sometimes outputs higher incorrect values which can mess up calibration causing it to sound low pitch and monotonic. This is usually uncommon, but if it happens you will need to reset and retry calibration or set your pitch mean manually.
- New Voices Over 100 original voices to try out and mix between for subscribers. Non-subs can try out 2. Also added are "Community Voices" which can be downloaded in-app. I am working on a pipeline to train new models which will be downloadable fairly quickly. Also working on a way for users to add their own voice data to train.
- In-app updates I should be able to offer new updates in the program, you will get a prompt asking to update and it should take care of the rest. This should let me push out updates easier, so you may see those more often than posts here.
- Free Voices. Non-patrons have access to 2 unique original voices. One male and one female. Also included are several Community Voices, though these ones are subject to change.
File conversion is not included in this release, along with many other things that are still being fixed, changed, or in development.
What I am working on now is improving the model to get the best possible quality. There are a lot of different parts to make it work right and I can't say how long it will take to get it sounding its best.
Important note. As mentioned above, there is a chance your pitch may not be calculated correctly. This will sound like a monotonic voice that sounds low-pitched. Either hit "Reset Calibration" and try again (with a normal speaking voice), or select "Details" and "Manual Pitch Values" then adjust your mean until the pitch line is around the middle.
If you are running into resource issues, there are a few things you can try to get it to run faster. Lower Chunk Crossfade to 15 and uncheck C1.
Important: It goes without saying, but at this early stage in development the program is prone to errors and issues. If you are looking for a production-ready program without issues, this is not for you. Additionally, I do not recommend subscribing until you test and see if NyVox works on your computer first and if the quality is to your liking.
Feel free to leave a comment or a message with ideas/suggestions/comments!
