The Process Forum

The Process Forum (http://www.process-productions.com/forum/index.php)
-   Transformation (http://www.process-productions.com/forum/forumdisplay.php?f=12)
-   -   Deep fakes (http://www.process-productions.com/forum/showthread.php?t=45941)

sybersamurai 09-04-2019 06:44 PM

Deep fakes
 
Just a cursory glance at the internet makes it seem this is still resource or time intensive at this point, but I know the quality is getting a lot better. Does anyone have any connections or know someone skilled in making "deepfakes"?
Was thinking would be awesome to see someone swap for instance Gal Gadot or Morena Baccarin into the American Werewolf transformation. Here's hoping.

TF-Viewer 09-04-2019 07:35 PM

Re: Deep fakes
 
It would be awesome if someone could pull that off. I've only ever seen the technology used in memes and porn though.

Anahki 09-05-2019 07:14 AM

Re: Deep fakes
 
just for wishing, my favourites would be Chloe Grace Moretz and Jennifer Lawrence, with special mentions to Maisie Williams and Emma Watson :P.

sybersamurai 09-05-2019 08:13 AM

Re: Deep fakes
 
Quote:

Originally Posted by Anahki (Post 828724)
just for wishing, my favourites would be Chloe Grace Moretz and Jennifer Lawrence, with special mentions to Maisie Williams and Emma Watson :P.

Great choices. There are some great concept shots of her from a dark Shadows. Also that movie has some good still shots of her transforming. Not great but solid for what it was.

Cursebearer 09-05-2019 11:15 AM

Re: Deep fakes
 
I expect this technology to get more prevalent in the future, potentially with tools powerful enough that even niche audiences like us will get what we want. The potential for virtual porn of all varieties is staggering. It's an exciting prospect!

Right now though I think we're still a little ways off.

mtsocano 09-05-2019 01:12 PM

Re: Deep fakes
 
I'm hoping that any of these becomes reality. We honestly need new female werewolves out there.

Garnu_Thorn 09-05-2019 03:23 PM

Re: Deep fakes
 
With my expertise in programming logic, I am planning on writing a collection of utilities bundled to streamline computer generated animations. The only barriers I have as of now is my predicaments with my situation in the mental health, lack of knowledge of modern programming language, and expertise in understanding of how animation of images and audio formation structure of the combined files actually work. Then there are also implications in what to do with said software as far as marketable feasibility and proprietary rights in the whole program and additions, patches, and subsequent other avenues where this software can be applied besides the entertainment industry.

slayer1 09-05-2019 04:39 PM

Re: Deep fakes
 
Quote:

Originally Posted by sybersamurai (Post 828725)
Great choices. There are some great concept shots of her from a dark Shadows. Also that movie has some good still shots of her transforming. Not great but solid for what it was.

Can you link these concept shots?

sybersamurai 09-05-2019 06:26 PM

Re: Deep fakes
 
Quote:

Originally Posted by slayer1 (Post 828743)
Can you link these concept shots?

Sure, here you go -
https://morphologyfx.com/pics/movies...olyn-wolf.html

Werewolf1361 09-08-2019 04:07 AM

Re: Deep fakes
 
Not to rain on the parade, but as this tech becomes more prevalent, it may raise some serious copyright issues.

odysseusut 09-08-2019 02:01 PM

Re: Deep fakes
 
I had a go at deepfakes just to see if I could swap one of my photomorphs into a scene, but the problem is that you need hundreds (preferably over 500) different photos taken from different angles and different lighting conditions, with not obstructions over the face (like hair, hands, clothing, etc) of each face that you want to swap, in order to train the AI. In order to swap face A in face B, they both need to stay the same throughout the scene. In that respect, I had way better results just using snapchat camera roll swap to put one of my photomorphs onto a scene, but that only works when the subject is forward facing the entire time.

Any sort of transformation would completely fall apart because face A (werewolf) would change throughout the scene making it impossible for the AI to learn what face A should look like, and make a complete mess of swapping face B in. You won't get the changes/transformations of face A carried over onto face B which is replacing it (the face of whoever you want to swap into the scene). The best case scenario is you'll get a normal unchanged human face on top of a werewolf body through the entire scene if it doesn't get completely messed up. You won't even get anything like contacts or teeth. The only thing that carries over from the original scene onto the new face is the facial expression.

TF-Viewer 09-08-2019 03:55 PM

Re: Deep fakes
 
Quote:

Originally Posted by odysseusut (Post 828855)
I had a go at deepfakes just to see if I could swap one of my photomorphs into a scene, but the problem is that you need hundreds (preferably over 500) different photos taken from different angles and different lighting conditions, with not obstructions over the face (like hair, hands, clothing, etc) of each face that you want to swap, in order to train the AI. In order to swap face A in face B, they both need to stay the same throughout the scene. In that respect, I had way better results just using snapchat camera roll swap to put one of my photomorphs onto a scene, but that only works when the subject is forward facing the entire time.

Any sort of transformation would completely fall apart because face A (werewolf) would change throughout the scene making it impossible for the AI to learn what face A should look like, and make a complete mess of swapping face B in. You won't get the changes/transformations of face A carried over onto face B which is replacing it (the face of whoever you want to swap into the scene). The best case scenario is you'll get a normal unchanged human face on top of a werewolf body through the entire scene if it doesn't get completely messed up. You won't even get anything like contacts or teeth. The only thing that carries over from the original scene onto the new face is the facial expression.

I see, yeah that makes sense why no one has done anything like it yet, it's just not possible. Oh well.


All times are GMT -7. The time now is 12:57 AM.

Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2024, vBulletin Solutions, Inc.