KevanOMG! I just improved Microsoft Sam. Took a block of text, running it through ElevenLabs, and reading the same text back. It sounds really, really good.
And now, enjoy Microsoft Mike and Microsoft Mary, both run through Eleven Labs. It takes these very old text-to-speech voices from 2000 and makes them sound a lot more human.
Want to hear Microsoft Mike run through Eleven Labs, but with an echo effect? The original had an echo, but Eleven Labs tries to remove it.
x0@kev OMG it's broken! The background sounds like FM radio interference or a noise reduction algorithm from a cheap laptop microphone gone wrong!
@x0 LOLOL it's also deeper than regular Mike was, check the thread if you wanna hear regular Mike, Mary and Sam.
🏳️🌈Superdave! 😘🍷👨🦯@kev That's a beautiful voice! Spookily natural-sounding!
Xantastic@kev oof, the reverb/echo thingy created all these interesting underneath artifacts.
@cordova5029 @kev I swear I've heard LPC codecs with too much noise on their input do that.
@cordova5029 @x0 Just put up TruVoice Peter.
@cordova5029 LOL check the rest of that thread to hear Mike Mary and Sam. Sam got greatly improved.
@kev OH I heard them yesterday, they're interesting. You can tell they're themselves, just way way optomized lol.
@kev go on, you know ya wanna try ATNT mike, or vw paul. How about VW/neospeech julie or James? lol.
@cordova5029 I don't have James. Maybe I'll try TruVoice. Oh yeah and I also tried eSpeak.
@kev I've heard espeak, delightfully broken :D
@cordova5029 I'd much rather have that inflection used by Eleven though
@kev OH same, that's not bad inflection. I bet it wouldn't know what to do with accented synth voices like Lee, Daniel etc.
@kev or, oh god, this would be positively gross, but l and h, peter and... oh what's the lady's name. Carol? She's the aamerican, I just forgot who the other l and h voices were. oh or true voice, that'd be nice and broken.
Austin Nix@kev that doesn’t sound a damn thing like even regular Mike. Looking forward to seeing what it does to Adult Male #1.
Trenton Matthews@kev sounds more like a stadium to me. Or, Mike's outside?
Morgan le Fay/Merlin@kev what would really baffle me is if somebody actually managed to make Microsoft Anna sound good with that thing
@evilcookies98 @kev While you’re waiting for an ElevenLabs version of Microsoft Anna, here!
Have an #UberDuck version of her for the time being!
@queenslight @evilcookies98 I can't do it, don't have Anna on this computer.
@kev holy shit ithat is weird. whaaat
@kev what does mike do.
@MariahL Not sure maybe someone will try it. About to post eSpeak's improved voice.
@kev Whoa! This is really cool!
@stirlock The site I used is https://elevenlabs.io.
@kev How do you upload your own voice to it?
@JesseF8693 You have to make an account. Voice cloning isn't free, 5 bucks a month for the basic level. But there's a free trial.
@kev Ah, so it's behind the paywall. Thanks. I might try that when I have a little more cash to spare.
@JesseF8693 At least it's not more expensive than that. But you get 30,000 characters only. There are more advanced plans though.
@kev Haven’t heard that voice in years. That is an improvement.
@kev If you’re going to say that sam sounds better than Mike and Mary, we need eleven labs versions of mike and Mary too. haha.
@austinnix Posted.
@kev laughing out loud . Good one!
@kev Oh my god i love it. rofl.
@kev That's awesome
@kev I, for one, welcome our new SAPI4 AI-generated overlords.
Kara Goldfinch@kev Wow. I wondered the other day if anyones done something like this. I wonder what dectalk or eloquence would sound like.
@KaraLG84 I did it with TruVoice Peter last night.
Pratik Patel@KaraLG84 @kev Here's my attempt at Eloquence from a few days ago. I haven't seen anyone doing this with Dectalk yet. https://mstdn.social/@ppatel/109831184285563118
Pratik Patel (@[email protected])
Attached: 1 audio Not sure if someone has done this before. Here's what happens when you give #ElevenLabs a sample of Eloquence to generate a voice. I wonder how it got some of the artifacts. Many screen reader users love the original voice. What do you think? #GenerativeAI #accessibility