AI Movies From A Talented Source
Steve K
Posts: 3,418
"Natasha Lyonne is making her feature-length directorial debut with an AI-powered film titled Uncanny Valley, a project from the AI-focused studio Asteria. The film, a sci-fi comedy written by Lyonne and Brit Marling, uses traditional filmmaking alongside cutting-edge AI tools to create an augmented reality video game world. The technology will be used for things like 'set extensions' and creating the immersive open-world game, rather than being an entirely AI-generated movie."
P.S. She is the star of the very good TV series' "Russian Doll" and "Poker Face" and a very intelligent, creative person. Also, "uncanny valley" describes the eerie feeling or sense of revulsion people experience when encountering humanoid objects or robots that look almost, but not entirely, like real humans.
So. Can we adapt this to Carrara animations? Is Wendy already doing this?
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
Sorry, I think as soon as you mention ai, I turn off. Ai certainly screwed my love of making digital art. Good luck to the woman.
I tend to agree, and I so far have little interest in getting into AI for any creative reason. Even though many of the hot Graphics cards are coming out in "AI" versions (?!) But if someone is going to use it in a productive way, I think she may be one of them. "Russian Doll" was a favorite TV series: "When the first season of Netflix’s 'Russian Doll' premiered in 2019, it took over my imagination for weeks. Its aesthetic was so specific and fresh, from the soundtrack to the makeup to the lighting." (RogerEbert.com) Lyonne is listed as both a creator and star.
This may upset a few people.
Consider using Carrara's vertex modeler to create a character in Carrara, rigged for movement and morphs for facial expressions. Even I have done it.
Consider using Carrara's dynamic hair modeler and Philemo's dynamic hair to mesh plugin to model a unique MESH hair style with movement morphs.
Consider using Carrara's native Toon filter tweaked with the GMIC render filters to create a non-photoreal look for the character (see that the dynamic hair to mesh plugin was necessary to render the hair with similar results as the renders of the character mesh).
Consider rendering roughly 20 images of your character in various poses and facial expressions, using your Toon/GMIC look.
Consider using the 20 Carrara renders in AI to train a LORA with your trigger word (Mychar?) which adds your character to the AI database.
Consider using the same Carrara Toon / GMIC combination to render a variety of your modeled props for a setting and objects.
Consider using the set/prop renders not to add those specific objects to the AI database (although you can), but to train a LORA for the render style so that the AI is trained for the 'look' you created in Carrara.
Consider using Carrara to render 6 key frames of a walk cycle of your character in your Toon/GMIC 'look', and with your depth maps from the multipass render list.
Consider loading your character key frames in AI, with your LORAs, including loading your depth maps in a Controlnet AI processor, and directing the AI to calculate the full walk cycle?
Now, consider using your character LORA with your style LORA, along with the Carnegie Melon BVH file motion database, Mixamo motions, etc. to create a variety of short stock animations of your character moving about in your look.
Consider combining the AI with your LORA's and your stock motions to create a 2-minute fake movie preview, or short happy birthday message for a friend, or short holiday greeting for family.
I reject the assertion that what I have described is theft. I reject the assertion what I have described is not art or has no soul. I reject the assertion that what I have described is any moralistic way worse for the world than using a 3D program to calculate the tweeners, or using Daz pre-built characters to create a 2-minute faux movie preview.
I am not 'there yet' in creating my non-photo-real movie preview with my characters in my look, but I am very close. And because Daz3D stopped updating Carrara, I will probably end up finishing my 3D-combined-with-AI journey in Blender and Filter Forge, not Carrara. In addition to everything else about Blender, the biggest issue for my dream project is updating the file format importer and exporters. For example, many motions can easily be acquired and used openly from Mixamo. Transferring from Mixamo to Blender is easy, but it is a bigger hassle to get the same motions from Mixamo to Carrara.
Note - Daz has updated its user license agreement. It would be problematic to use 20 straight Iray renders of Victoria 8 or Michael 9 or... to train a LORA for commercial use, or another way that competes with Daz3D.
I mentioned modeling a character in Carrara. If you have such a character mesh, here is how to upload it to Mixamo, and then export it for animation in Blender.
I happily use both
I however understand and sympathise with Headwax and others affected by Ai
I do try to keep my mentioning of Ai to stuff involving Carrara or other DAZ rendered content on this forum, I mostly use it for postwork and backgrounds with 3D or animations of my renders
I mostly only train LoRAs on real people and cats, the exception being an Ai character I refined based on a combination of men I find hot.
I always thought training LoRAs on DAZ content would be derivative
I wouldn't train one on other peoples art either unless in the public domain and wish they had a cleaner database for generative Ai
I do find many videos by others using it fascinating and see it as an art form in itself
Yes, I agree. I sympathize with everyone affected. And, to that end, I have a thread in the Art Studio forum where I have moved most of my AI-related comments, questions, and discussions, unless there is a direct relation to Carrara, such as the multipasses that produce depth maps, outlines, etc.
Furthermore, I am particularly upset by genuine theft. We have rules for personal property rights, rules for public domain, and rules for licensing. In addition, we have harassment rules. People who download 20 pictures of Taylor Swift or Emma Watson, etc and then train a LORA and then create porn which they derive income from view-counts are violating all sorts of laws (plural) and ethical norms (plural). They should be pursued, deprived of the income they generated even though they did not charge a membership fee, and prosecuted for harassment. At a lesser evil end of the wrongful spectrum, there are people who created AI databases under false pretenses, such as the people who promised to be non-profit when they asked the GettyMuseum for their photograph collection for free, then turned around and competed with Getty for access to the pictures or derivatives. Not as abusive as what Emma Watson endures, but still vile and should be pursued to appropriate redress as determined by Getty.
But you can see I am also triggered by people who paint with a broad brush and say EVERYTHING done with AI is theft. No, we have a public domain with rules for the public domain for good reason. If I go back to the original source material and make a Frankenstein video with the character I created and generated a LORA for, it isn't theft. Not even remotely. And in the US, I can legally use my LORA to remake the 1924 classic Thief of Baghdad even using AI motion capture on the original movie to aid the animation. That film is truly in the public domain under US law. What I would be doing would not be theft.
Recall that the English film company Hammer went back to the original Frankenstein material and made their Frankenstein in the 1950s even though the US film production company Universal made a 1930s version. The 1930s version was NOT in the public domain, but the source material in Mary Shelley's book was. So Hammer could, without committing theft, make their own Frankenstein as long as they used their own music (or music they licensed the rights to), their own art design, their own style props, their own style costumes, etc. Shakespeare is in the public domain, obviously.
In contrast, the people who made the film Airplane did go back and pay for the rights to a movie called Zero Hour (I believe Canadian production). The point is we have rules and plenty of people try to abide by the rules. There can be disputes even when everyone acts in good faith. But I get triggered when people call something theft even though the facts available to the public are that the creator used public domain material, or material for which they obtained licensed use, or used the material in such a way that it met the standards of fair use.
Repeat, the producers of Airplane! did pay for the rights to Zero Hour.
Oh no need for sympathy. Thanks. I do other stuff now. ;) of course I speak to alexa who is stupid ai.
Edit: snipped the rest of my remarks
Have personally seen both sides of AI. A person who I know was willing to pay for 3D renders in his interest area, I did some renders for him then charged him what it cost me in paper & DAZ models I had to buy for him. Now he.s on the AI gravy train and refuses to pay for renders.
For myself I use it to generate concept art eg Aliens & Spaceships. I have also now started to translate foreign magazines, scan a document PDF or png into an OCR, then copy the text that is generated to AI and get it to translate it. Translated a heap of French Aviation articles the other night.
Sorry for hijaking a thread once again !!!
Bachman Turner Overdrive - You Ain't Seen Nothing Yet 1974 Video Sound HQ
love that song.
Interesting ideas in this thread:
I guess it;s a cultural thing:
In Oz we think that there is a difference between legal and moral; we also have this far fetched idea in that, just because it has been done before ,does not make it right.
"it's okay for me to do it because someone else has done it", isn't a wonderful argument. In fact, it's a basic fallacy they teach you in week one, of 'how to think logically, 101'
.
As an aside, I like the idea of training ai on our own work.
Of course, to make it moral, you would have to write the ai program yourself to make sure it wasn't trained on stolen work, or work used without permission.
I'm sure most artists saying they only train ai with their own work would be aware of this.
Legality versus morality and honesty - what a difficult choice.
Well there is FAR more going on with the AI discussion than most realize. It will be redefining many things. However, one thing is certain it IS and will be everywhere. AI has been around for a very long time. It is just now being publicly pushed because it is key to massive commercial markets in the billions and trillions.
https://en.wikipedia.org/wiki/ELIZA_effect
https://en.wikipedia.org/wiki/Artificial_intelligence
Quote:
Artificial intelligence was founded as an academic discipline in 1956,[6] and the field went through multiple cycles of optimism throughout its history,
great song... performed originally by Randy Bachman as a joke for his stuttering brother...
Taking Care of Business by Bachman Turner Overdrive was another hit when during recording rehersal the pizza delivery fella ( who delivered some pizzas ) said he would fill in playing the piano... then after he went back to work at the pizza shop, there was a mad scramble to find him again to get his permission to release the song with his piano playing
brit marling is the creative talent there, not natasha lyonne, but that aside...
i think intentionally using AI in full knowledge of its dystopian aspects and to elicit revulsion is the only justifiable means of using AI
This musician has some interesting demonstrations and discussions of using AI in the music industry, for example.
Another good one Stezza. Regarding, BTO.
This is essentially why I started focusing on Blender. Many options there. Blender is becoming very robust.
The thread opener is a bit of a narrow topic so it is bound to go in other directions. The production mentioned makes it clear AI is being used. This example shows problems created with out this kind of clarity.
Thought I would update this. Apparently, this AI channel is not only violating YT TOS because it is undeclared, but the owner of said AI channel has had "his" AI music stolen, posted and registered on music sites for money by others. This is how this can spiral.
Another example of AI rules in a current still image contest:
"AI Usage:
Not Allowed: Any image generated wholly by an AI seed or keyword prompt.
Allowed: Artist-created images with AI used only for postprocessing (color correction, tone, saturation, or special effects).
Requirement: Provide a step-by-step workflow in your description. Renderosity may request WIP screenshots or other proof; failure to comply can lead to disqualification.
"
Fwoof. Glad I'm not using AI, sounds like a lot of "step-by-step workflow" paperwork (so then I moved the character a little to the right, and ... ).
I think that rule is an attempt to be fair. I also think that with the amount of AI built directly in image editing programs these days it is going to be difficult to maintain a strict 'No AI of any kind rule.' I tend to do a lot of screen-shot side-by-sides anyway, so the 'document what you did' mandate wouldn't bother me much. I could see how it would discourage a lot of people, though. Remember the kitten was 25 years ago! (watch linked video)
Yeah sadly, that AI channel seemed to think he was actually song writing. This is not just about ownership, but about aesthetics, process and ethics. Yes AI is everywhere and it will only become more pervasive. It is interesting though how it will redefine and challenge Values.