Re: microsoft-espeak-forked tts
@DracoSelene89 That's precisely what I'm saying. Microsoft hates open source software yet they pretty much welcome it into windows 10. How do you explain that? And Ray, ok I'll agree with you on that. As much as post 11 was a slam on the prerecorded samples, I'll agree that the voices themselves definitely sound good for what they are. Responsive? No, unless you're using the mobile voices, which definitely do sound better and are also available directly through an nvda addon that exposes nvda to the mentioned exploit to get the voices. The latency and the prerecorded samples are pretty much the only problems I have. with the voices. I just think the prerecorded samples are just not the way to go for synth creation, because 1 they don't support pitch/speed change, or at least not as effective as most synths. It's like a non-server dependent version of speechhub, in a sense that the most used phrases are essentially loaded into memory. Now don' ;t get me wrong, I like Speech Hub, but you'll notice in some instances, if you change the pitch and the speed of a voice after hitting some keys, next time you hit those keys you may be getting the samples you heard before. That's what I'm talking about when I say it's all preloaded into memory. Except for windows 10, that's always how it is, they're just loaded into the framework, permanently. The actual synthesis units are more than satisfactory, but the samples, not so much. Then again, tts is not Microsoft's department, and most probably wouldn't care in the least. Which is why I do agree, Microsoft partnering with Nuance would be a good idea despite Nuance's track record of borderline unethical domination of the market, but it would really give Microsoft a solid synthesizer to work with, allowing them to focus their efforts on accessibility without worrying about the speech implementation, so long as the implementation was just as s olid. And no matter how much you try, just know that espeak will never sound like eloquence. You may get the algorithm if you try hard enough, but keep in mind that most parametric synths use vocoders. Don't believe me? Turn some of the dials down to 0 on the nv speech player, and you'll notice you'll hear straight tones. That's exactly how it works. The phonemes are ran through signal processing and matched to the vocoder's processing as well. Normal speech would be samples as the modulator with tones as the carrier. A whisper voice consists of the same phonemes as the modulator, and a white-noise waveform as the carrier. Eloquence's signal processing is different to that of other synths, so even if you managed to get the algorithm pitch perfect, you'd never be able to imitate the true sound of Eloquence due to the different signal processing. It's akin to trying to make Espeak sound like a Votrax.
_______________________________________________ Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
