This is a really, really bad idea. But the way Dengjen Neural Voices can pan the text to speech audio, based on the on-screen position of the object being read, is so cool! And NVDA addons can patch NVWave. And unspoken-ng is a 3d audio addon that I already maintain. Man, addons shouldn't do this. Addons shouldn't be allowed to do this. But they can do this! This is going to break so many things. This feature shouldn't exist. But it's so neat! Someone convince me not to do this. #nvdasr
@fastfinge oh boy. Imagine someone's add-on patching NVWave and then another Speech synthesizer not working as expected for users, then the author of that other synth gets the blame. Oof. Definitely one to be careful with, but I do get the enthusiasm, truly. Never realized those neural voices could pan based on the object itself, very neat.
@Tamasg Yup, they can. But...I can do it better in unspoken-ng, for every synth. I use openal for HRTF stuff. I can just grab the audio any synth generates and shove it through my existing pipeline that I use for unspoken-ng's earcons. Then the voice can come from the same place as the earcon, perfectly in sync. For any synth that sends audio through NVWave, anyway. The real reason this is a bad idea is because once I do it, someone else is also going to want to do it. What happens when someone installs two different addons, and they both are messing around with NVWave? Bad things happen, that's what happens.
@fastfinge oooh no kidding this is neat. I suppose one option is if an addon detects others that could patch NVWave player. It does mean awareness across developers and etiket to do that for users, but yeah. Whether through a list of other "blacklisted" addons it cannot work alongside or similar. If the addon has a way to disable the patch from applied, overwrite the setting. Since it's a brand new concept the only good news is not many on that list exist yet. Ha.
@Tamasg I have it set to do nothing by default. It only patches if the user changes the setting to ask for panning or HRTF. So by default, just installing the addon and using it as normal won't break anything. And if it can't patch NVWave, it will fall back to not doing anything, so users shouldn't lose speech. In theory.
@fastfinge yeah that's a clever idea too! It's working really well here, even with other synths I threw at it like SSI263. One downside: If a page alert comes in, or any other announcement through NVDA Helper, it stays positioned at the last spoken object's position rather than center-panning. This is a very minor one considering how quick you got it working :D
@Tamasg I'm not actually sure how I could fix that. When speech is generated, NVDA doesn't give me a way to know if it's from something like a system notification, rather than because the user is tabbing around or something. You can know for say-all, but as far as I'm aware, that's about it.
