TresCom Forums

Yeah this is exactly why I don't feel so comfortable mixing this AI voice with his real one. That and the sound quality still isn't quite there. But like the creator of this AI said, give it a few years and maybe it could...

_________________

It essentially falls in line with the ethics side of things. People that clone voices on youtube usually declare somewhere in the description that the voice is a clone and they credit the original actor. They also usually link to the TacoTron 2 information page. To my knowledge, there hasn't been any takedowns or legal ramifications regarding the cloning of voices. (god help them once the technological singularity gets hit circa 2045-2050 cause once that happens, the voices and highly likely far more will literally be cloned themselves by super advanced AI's far superior to what we have now in minutes)

I find these AI voices to be quite interesting, even if they're not quite there in terms of realism. Wouldn't consider them quite high-quality enough yet to be of use for much, though.
I do agree with the ethics debate - I recall seeing a video where they got Shaggy from Scooby-doo to say some things that Casey Kasem would never have said via one of these AIs. I don't have this problem with voices which are made specifically by actors for the purpose of being used for these AIs, but I find it hard to believe that these "learning" voices that copy other people without their permission would go over well with many of them.

_________________
Current Project: Cloning Facility | My Release Thread

So, the people have spoken on the Twitter poll, and a bonus track(s) will be made instead of redoing the entire memoir! As I mentioned earlier in the thread, I agree that this is the best option, at least for the time being. A new edition just came out last year, and maybe in a couple years time the voice tech could improve and mesh better!

Out of 43 votes, 41.9% wanted a redo, whereas 58.1% only wanted a bonus track. So that's still a large number of people that wanted a redo, definitely keeping it in the realm of future possibility. But sorry guys, this election wasn't rigged, so we gotta move on.

_________________

from what i've heard in the youtube link, there is a similarity in some aspects, but there are differences that are really making it easy to discern that it has been synthesized. The diction is ok, but there is a tonality that remains perfectly unchanged in his depiction of kills and even when sitting at a crappy business laptop you can hear the difference between audio oscillations from his real recording and the synth version. besides, other than for humorous attempt, the character of hammond doesnt seem to be having such malicious conversational skills. He has just been shown as too jolly and childlike to be giving these lines. i cant put my finger exactly on the term, but i'd say that there is a note that the voice doesnt drop below and it makes the clone version noticeable as such.

Yeah the discernable difference still is why I am glad the majority of people would just want me to assemble his unrecorded lines separate from the recorded ones. At this current time it will still stick out a bit. But I disagree that the voice doesn't come off with Hammond's tone. It VERY much does match his tone from Trespasser since the recordings used are literally from them. I have always loved this more somber take on his character that is present at times in both Jurassic Park and The Lost World, but I think most people think of his spirited "Welcome To Jurassic Park!" style of talking VS something like when he talks about "Petticoat Lane". Trespasser is like 45 minutes of purely "Petticoat Lane". So while it doesn't capture that spirited part of his personality much, it's still authentic to a point.

_________________

This is really cool.

I spent some time reading some lines i had planned for my project and i've got to say its a bit scary how accurate they sound. You need to tweak the spelling a bit to capture his unique pronunciation for certain words, but man this is going to be handy one day!

_________________

As I said in the other topic, some amazing stuff

_________________
Active project: Trespasser: Isla Sorna - Working on PH2-IT (and experimental BE-IT).

"...there used to be more benches, but InGen's workers removed them during the evacuation in the name of framerate."

All the new tracks are up. Check out the site for the full playlist, the new audio download, and the new book download.

https://jurassictime.wixsite.com/memoirs/hammond-memoir

End of an era.

_________________

For anyone that previously had trouble using the Google Colab workarea, I have now submitted this to UberDuck.AI whom have accepted it on the site.
https://uberduck.ai/#voice=john-parker-hammond

It's literally as simple as typing in what you want then hitting the synthesize button

OMG, the AI is insane! It even uses Attenborough's breath while pausing when using punctuation. And there seems to be no more vocal fry in the voice now

Is this UberDuck AI better than the other one? I'd like to test it out but it seems you can't log in without a Discord account.

_________________
Current Project: Cloning Facility | My Release Thread

Two tests to demonstrate differences:

The original Colab:
https://drive.google.com/file/d/1Pmfwf7 ... sp=sharing

The UberDuck.AI version:
https://drive.google.com/file/d/1qXgJrw ... sp=sharing

The UberDuck.AI version shows a massive improvement in terms of how cleaner and how much fuller the audio sounds. As such, I'd strongly recommend use of the Uberduck.AI version for anyone that wishes to produce any content using the John Parker Hammond voice. The difference is literally night and day. The worst of the "electronic" phasing that you hear on the original colab version doesn't appear to happen in the uberduck version.

Do you think you can get Anne's AI on there as well? There's some lines I want to create for a custom level