AI Has Turned Me into a Gambler

Nyghtfall3DNyghtfall3D Posts: 777
edited May 5 in The Commons

Does anyone else currently experimenting with AI feel like you spend more time pressing the Generate button in the hope of seeing something that "works" instead of making deliberate progress creating something you intended? I've spent entire evenings trying to work out small details in projects I could've wrapped up in a few minutes, and it's really starting to bug me. This morning I found myself asking, "Is AI photorealism really worth all this hassle?"

Someone on another board I frequent once said they like to think of AI as a type of "render engine". I wish it actually worked like one. :P

I want to be content with Iray again...

Post edited by Nyghtfall3D on

Comments

  • LinwellyLinwelly Posts: 5,956

    I quickly lost interest after experimenting for a bit, I might still use it for a background at times, but I I know that with a bit of work I can create what I really want with DAZ studio, so much more satsfactory

  • MimicMollyMimicMolly Posts: 2,200
    My problem is that I like to edit/paint over whatever decent AI result I get. But sometimes it feels like it would've been a better use of my time to simply draw my illustration from scratch, since I'm fixing too many mistakes.
  • FSMCDesignsFSMCDesigns Posts: 12,755

    Nyghtfall3D said:

    Does anyone else currently experimenting with AI feel like you spend more time pressing the Generate button in the hope of seeing something that "works" instead of making deliberate progress creating something you intended? I've spent entire evenings trying to work out small details in projects I could've wrapped up in a few minutes, and it's really starting to bug me. This morning I found myself asking, "Is AI photorealism really worth all this hassle?"

    Someone on another board I frequent once said they like to think of AI as a type of "render engine". I wish it actually worked like one. :P

    I want to be content with Iray again...

    LOL, I feel ya. Since my vacation ends today, I have been going thru the 200 styles in a certain AI program to see what results they would give. You have an idea what to expect based on the prompts, but each style does so much more that just change the artistic look (like a image editor filter), they redefine models, shapes, poses, it's quite amazing on one hand and frustrating on the other when you don't get exactly what you expect, LOL

    I doubt I will ever be content with Iray again, just because of the sheer variety of options that you don't get with DS, or without having a TON of differnet PAs pumping out variations on outfits daily. My test prompt on an earlier test of styles was "Red Sonja" and the sheer amount of sweet looking variations on her wardrobe alone made me wish they were avialible in 3D. Hopefully there are some PAs using AI for inspiration.

  • paulawp (marahzen)paulawp (marahzen) Posts: 1,396

    I spent 15-20 minutes one night trying to generate a nighttime street scene to be seen through a small window in a room in a traditional Daz scene, rather than spend a few minutes loading and positioning a "real" street and just getting what I wanted. The only reason I tried was because I thought, "Since the details don't matter, and just a small bit of it is being seen through this window, surely it'll be faster to get a reasonable, basic street scene in AI, than loading one in Daz." 

    Nope.

    I got so disgusted by wasting my time on something so elementary that I quit bothering with the original image and went to do something else.

  • linvanchenelinvanchene Posts: 1,382
    edited May 6

    I actually got bored of always knowing what an image will look like when setting up a 3D scene.

    I started to put together random images with layer elements like background, clothing and hairstyles selected from python lists.

    With wildcards, dynamic prompts and AI Generators I can get random results a lot faster now.

     

    I do agree with the sentiment that I am now more an image curator than a creator.

    I look at a sample of 8 images. Download the best ones. If needed I reroll and download the best ones again.

    From the downloaded images a selection is chosen for postproduction and publication.

    Post edited by linvanchene on
  • WendyLuvsCatzWendyLuvsCatz Posts: 38,289

    I say I am a Producer

    I use AI in videos for images, animation, special effects, music etc

    a movie producer does this, they don't usually act, write/script, build sets etc

    I do use 3D and other tools too

  • nonesuch00nonesuch00 Posts: 18,163

    I tried to 'fine tune' control some and it did, a little bit but not enough to suit me by a long shot so I just put in a string of text and see what it does. Sometimes they are are the verge of being very nice but most often that are severely lacking in certain areas. Because of that, I only generate images occasionally, preferring to accumulte credit for when the results are more consistently anatomically correct or I can more finely control the content, or both even. Also I expect credits to buy more later on.

  • Bunyip02Bunyip02 Posts: 8,691

    I am still doing renders in Carrara, even though I have spent more time on AI Imagery than I thought I would.

    Am using the AI to broaden my thinking about the possibilities that I can extend my Carrara renders to.

    Is another tool/plugin to use for my renders.

    Also sharpening up my 3D modeling skills as I now have lots to model.

  • FSMCDesignsFSMCDesigns Posts: 12,755

    linvanchene said:

    I actually got bored of always knowing what an image will look like when setting up a 3D scene.

    I started to put together random images with layer elements like background, clothing and hairstyles selected from python lists.

    With wildcards, dynamic prompts and AI Generators I can get random results a lot faster now.

     

    I do agree with the sentiment that I am now more an image curator than a creator.

    I look at a sample of 8 images. Download the best ones. If needed I reroll and download the best ones again.

    From the downloaded images a selection is chosen for postproduction and publication.

    based on the posts i have seen by you, you are a seasoned and experienced user with AI tools, language, structure and concepts, something many are not or ever will be. It's one thing for a casual user like most here on the forums to see a generator, put in some prompts and get a cool image, but it's another for these same users to know what a LoRA, or check point is, how they work or even how to create one They would also be lost if they saw the UI of ComfyUI or Automatic1111 or had to figuire out what version of SD they need for the AI tool they are using. My point being that while AI has come a long way and now has the ability to reach so many more users because of easier UI's like what DAZ has and online tools like ImagineArt AI, there is still a big disconnect between random image generation and getting the image you actually want with AI tools with casual users which is the point the OP was trying to make. Sitting there clicking generate in hopes you get something close to what you actually want, isn't very productive, fun, but not productive.

    Very experienced AI users know how to get close, if not exactly the results they want with AI, just like experienced DS users know how to get the results they want with DS and iray, so we still have a ways to go if anyone wants the DS crowd to embrace AI since that control is missing. Right now it is mostly random for the casual user.

  • WillowRavenWillowRaven Posts: 3,787

    I think Daz's foray into AI would work better if I could show the AI a photo of a person and DazAI shows me the closest Daz products to that person so I can just buy that character rather than spending hours searching through all of Daz's characters from V4-up to get the figure I want. Or show  DazAI a photo of a dress or outfit or shoes, and it spits out a list of products closest to the photo. That would save me more time and money than random, scraped images never giving me what I want.

  • TorquinoxTorquinox Posts: 3,366
    edited May 6

    WillowRaven That would be cool!

    Post edited by Torquinox on
  • WillowRavenWillowRaven Posts: 3,787

    I spend so much time trying to find what I need for a commissioned scene; if DazAI did THAT, I could spend my time creating my scene. I'd get more work done faster, meaning more commissions, meaning more money and PAs getting paid. How does this crap help the PAs? Or am I totally misunderstanding the About page?

  • FirstBastionFirstBastion Posts: 7,771

    Daz AI will probably evolve over time. The direction though, is the gamble.

    1st-roulette.jpg
    750 x 750 - 104K
  • WillowRavenWillowRaven Posts: 3,787

    Until it's useful to me, I don't see me bothering with it. 

    Question ... How does it help the PAs sell more products?

  • Ghosty12Ghosty12 Posts: 2,060
    edited May 8

    My big concern is with what is going on will Daz favour doing more work on its AI, and forget about Daz Studio 5 altogether. Since really this AI thing pretty much came out of left field.

    Post edited by Ghosty12 on
  • Nyghtfall3DNyghtfall3D Posts: 777
    edited May 10

    Not that I mind, but it looks like I'm stickin' with Iray for long-form comics.

    I've been spending part of my vacation this week re-organizing my Daz content library and decided to see what I have to look forward to with learning how to create AI enhanced photoreal comics while maintaining character consistency across panels. Suffice it to say, it doesn't look possible with AI's current level of development.

    I'll keep using AI for one-offs when possible.

    Post edited by Nyghtfall3D on
  • ArtiniArtini Posts: 9,503
    edited May 10

    For me Daz AI Studio provides so much fun and provides ideas for the further development.

    I could not imagine any other faster way to test my ideas for so broad range of visuals.

     

    Post edited by Artini on
  • 3DmentiaNull3DmentiaNull Posts: 132

          So far I've only used Daz AI (free version), and, "push da button" is certainly addictive. But seeing the small image on my laptop is usually deceptive and when seeing it larger so many errors become apparent. You know, the usual stuff. So one of the impressions I get is that the AI is not starting at the back of an image and working forward. It's as if there is a decision of what elements are to be put in, then it's looking at the scene and getting confused as to where things are positioned in 3D space. Anyway, the most fun with it has been to see it create "M.C. Escher" like effects in mundane scenes. surprise Maybe if I put in a prompt for a M. C. Escher theme room it will come out totally normal. laugh AI is also artificial ignorance smiley / peace

  • WonderlandWonderland Posts: 6,944

    WillowRaven said:

    I think Daz's foray into AI would work better if I could show the AI a photo of a person and DazAI shows me the closest Daz products to that person so I can just buy that character rather than spending hours searching through all of Daz's characters from V4-up to get the figure I want. Or show  DazAI a photo of a dress or outfit or shoes, and it spits out a list of products closest to the photo. That would save me more time and money than random, scraped images never giving me what I want.

    Apparently Google will be able to do that soon. They are ramping up their AI right now and Google will soon have a completely new look involving AI. 

Sign In or Register to comment.