Open nakazax opened 1 year ago
Hi @nakazax , thank you for reporting an issue🥰
I couldn't notice this issue because it looks working well with ja-JP-HarukaRUS
in my environment.
I will change the out-of-the-box default speaker name in the next update.
Will you give me your personal recommendation for the speaker? (ja-JP, female)
Hi @uezo, thanks for your comment. The Azure document says
you used a standard voice with your Speech resource prior to September 1, 2021 then you can continue to do so until August 31, 2024. All other Speech resources can only use prebuilt neural voices.
It is likely that you were unaffected by this if you had been utilizing a speech service that was created prior to July 2021.
My personal recommendation is ja-JP-NanamiNeural
. It's the most standard female voice, with no noticeable quirks, making it easy to understand. You can listen to voice samples of this and other voices at the Voice Gallery.
Hi @uezo, I created demo videos of ChatdollKit and Azure TTS neural voices, so I'll share the links with you.
ja-JP-NanamiNeural
https://clipchamp.com/watch/ahhggY8LeG6ja-JP-AoiNeural
https://clipchamp.com/watch/eEVdZp3znsxAfter listening to the demos, I think the ja-JP-AoiNeural
voice may match better with this 3D character.
P.S. I created a C# script using Azure OpenAI Services' ChatGPT as an AI engine and used it. I'll send a Pull Request later separately from this issue. Thanks!
Hi @nakazax ,
Thank you so much! Okay, ja-JP-AoiNeural
is the next OOTB default.
Currently, Azure TTS Loader is using the
ja-JP-HarukaRUS
voice, which is one of the pre-built standard voices. However, this voice has been retired, resulting in a 403 error when attempting to use it. To resolve this issue, we need to switch to a Neural TTS voice, such asja-JP-NanamiNeural
. You can find more information about Neural TTS voices at the following links:Error message