I think this is just using SpeechDispatcher from the system -- so it's not a Firefox specific thing. I get a similar (but very slightly different) voice on my own system by default -- which matches what I get when I run a command like spd-say --wait "Hello world" from the command line.
I'm pretty sure SpeechDispatcher can be configured to use a different synthesis engine -- Arch's wiki has some suggestions: https://wiki.archlinux.org/title/Speech_dispatcher -- but I haven't dug into it yet.
Hm, ok interesting, then I could appearently use Piper (from HomeAssistant) which has a nice sounding voice, I'll try that when I have some time: https://github.com/rhasspy/piper/discussions/328
Yeah, in my opinion this sounds worse than DECtalk from all the way back in 1983 (for instance, Perfect Paul). Or MacinTalk (see its usage in Whatever Happened to Robot Jones, particularly the original version).
The newer stuff often tries to sound less robotic but then if it doesn't have the complexity to sound realistic it just sounds worse, harder to understand and yet also sounding uglier... warbly.
Depends what you mean by newer stuff. OpenAI is pretty good in terms of tts
I meant newer than the context I stated (1983). And also ground floor stuff that doesn't need internet (after install) and doesn't need a purchase/sign-up. Available by default or not too obscure to get decent voices.
There might be some half-decent voices somewhere, but it really just doesn't seem like it's night-and-day for the ground floor stuff from what I've seen. Maybe some vocaloid stuff but even that seems like a chore to do the phonemes manually to get expected pronunciation.
Why are these voices a feature of the browser instead of the OS?
I think this is just using SpeechDispatcher from the system -- so it's not a Firefox specific thing. I get a similar (but very slightly different) voice on my own system by default -- which matches what I get when I run a command like
spd-say --wait "Hello world"from the command line.I'm pretty sure SpeechDispatcher can be configured to use a different synthesis engine -- Arch's wiki has some suggestions: https://wiki.archlinux.org/title/Speech_dispatcher -- but I haven't dug into it yet.
Hm, ok interesting, then I could appearently use Piper (from HomeAssistant) which has a nice sounding voice, I'll try that when I have some time: https://github.com/rhasspy/piper/discussions/328
Yeah, in my opinion this sounds worse than DECtalk from all the way back in 1983 (for instance, Perfect Paul). Or MacinTalk (see its usage in Whatever Happened to Robot Jones, particularly the original version).
The newer stuff often tries to sound less robotic but then if it doesn't have the complexity to sound realistic it just sounds worse, harder to understand and yet also sounding uglier... warbly.
Depends what you mean by newer stuff. OpenAI is pretty good in terms of tts
I meant newer than the context I stated (1983). And also ground floor stuff that doesn't need internet (after install) and doesn't need a purchase/sign-up. Available by default or not too obscure to get decent voices.
There might be some half-decent voices somewhere, but it really just doesn't seem like it's night-and-day for the ground floor stuff from what I've seen. Maybe some vocaloid stuff but even that seems like a chore to do the phonemes manually to get expected pronunciation.
Why are these voices a feature of the browser instead of the OS?
They are! OP made an error in their post.
The best part is when it fucks up the HDMI audio output of your expensive graphics card.
Yes, I've experienced the same issue as the guy who posted the solution.