nitpicky about how smoothly speech synths can speed up

Hi all,
So this interesting thought came to me earlier. Admittedly I've been musing on it for some time.

As a perfectionist/audiophile, I pay attention to details most people would either find unnoticeable or nitpicky, and what's worse, I often let said details bother me to no end.

So here's where I am currently. I use Eloquence full time. Right now i'm using IBMTTS for NVDA. I personally think Eloquence is great. It's been my main synth exclusively for the past 20 ish years. I tend to use a wide range of rates. I always have rate boost on, but I will change the speed a lot.

I stick to around 20-25 percent when reading without any real concentration. This is a rate many sighted people consider quite fast. For me, it's just above natural speech rate. For messaging or general computer use, I hang around 30 to 40, depending on how alert I am, and I can go up to around 50-55 for skimming if I concentrate. It might not sound like a wide range, but every 20 percent doubles the speed pretty much, so 55 is roughly 3.5x the speed of 2.

But my goal here isn't to slobber on Eloquence and revive the "We must have Eloquence in our lives" discussions. Actually I'm more wondering what I would do if I lost it tomorrow. I'm not necessarily preparing to stop using it, but I'm just idly curious. I have not been able to, with confidence anyway, find a synthesizer that can nicely accommodate the rates I would like to use, and i'm wondering if this is a common thing.

I've tried ESpeak with rate boost and, while I could certainly get used to it, it really does bother me. Around 10-15 percent with rate boost is my comfortable reading speed, and I could get used to that no problem if I had to. I have fairly weird ESpeak settings which only sound nice to me on my favorite headphones (Klat4 with american English), but that's beside the point. Anyway, 10-15 is comfortable reading, 25-30 is normal casual rate for messaging/computer use, and skimming would be maybe 50? I'm not really used to it enough to say for sure, but it's something like that. I know some poeple crank it up into the 60s, maybe 70s though, so people are going faster than me and are doing it regularly by the sound of it. So yeah, their requirements would be even firmer than mine.

My problem with ESpeak though is that after around 15 percent with rate boost on, male voices especially have this horrible croaky quality. Even with rate boost off, 100 percent starts to show it a little. Others have told me this doesn't bother them, so I'm sort of wondering if I've been majorly spoiled by Eloquence. Also, ESpeaks' consonants aren't quite right to me. So the speech either sounds jerky to me because they're too long, or because they're too short. Or maybe it's not so much the length but the transitions. In all fairness, it sort of balances out for me after the rate speeds up a bit, but there's a point right in the middle of my comfortable reading rate range where consonants and croakiness really bug me. Admittedly, for really fast rates, all of my qualms with it might possibly be imperceptible but still. I don't routinely use those rates.

The Windows OneCore voices aren't too bad either. I feel like their pauses are too long for me, so I'd have to get used to that. And they sound a bit odd when speeding up to extremes, but I could make them work if I had to. I think personally I would be switching back and forth between ESpeak and OneCore. When I got tired of one I'd switch to the other most likely, not really feeling satisfied, at least not at first.

All three of the above synthesizers can go faster than I need (OneCore's fastest is on the outer limits of what I can skim through so I can understand it, only just). I commend them all for that. Other synths can't claim that though. I've tried Vocalizer, especially on IOS, and demos of Ivona and Infovox. They didn't seem to go fast enough, I was routinely pushing them to 70 or 80 or something like that, and some really sounded bad when I did that. Admittedly, for normal reading they'd work just fine, and for moderate speed-up they work too, at least some of them, but for skimming and quick responsiveness, I know they would slow me down, some more than others, if I had to rely on them.

The best I've found is, oddly enough, on IOS. Not Fred... he's insanely responsive but is possibly the worst speeding up voice when trying to use the rates I like. Instead, I go to the US Siri male voice. It sounds smooth to me when speeding up. I sometimes hear artifacts, but rarely notice them. No croaky sound either, I love it! So, that is what I use on IOS now. People have made fun of me for using the gay gymnast voice, and I can see sort of where they're coming from, but the Siri female voice is outright cringy for me personally so I can't stand it. But at 70 percent I think the Siri male is crisp and clear and is what I routinely use. It can't get to my full steam skimming rates, but 100 percent is starting to touch that. on headphones at least, I could use 100 for a moderately long period of time, although it's starting to struggle a little. So yeah it's not perfect, but all in all I'm in favor. Of course, I couldn't use it on PC because it's exclusively for IOS and perhaps Mac (I don't have a Mac to confirm that). So yeah it doesn't help in the grand scheme of things, but at least I don't have to fight to find a favorite voice on IOS anymore. If I could put it on PC, it would be a very strong contendor for an Eloquence replacement, at least for the majority of things, believe it or not. The two are vastly different, but my comfort level when listening to them is weirdly similar. The Siri male may lose some points on responsiveness, though.

So I guess my point is this: the ease at which Eloquence speeds up is a big reason, though not necessarily the only reason, that I've stayed with it. Can this be justified objectively, or is my bias and or pickiness clearly in evidence and that's the end of it? Any weird insight to be drawn from this perhaps?

-- 
Audiogames-reflector mailing list
Audiogames-reflector@sabahattin-gucukoglu.com
https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
  • ... AudioGames . net Forum — Off-topic room : musicalman via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : musicalman via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : Turret via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : musicalman via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : GrannyCheeseWheel via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : PatrykK via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : burak via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : burak via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : nidza07 via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : musicalman via Audiogames-reflector
    • ... AudioGames . net Forum — Off-topic room : haily_merry via Audiogames-reflector

Reply via email to