SyntaxBomb - Indie Coders

hi ! 

i have played a little with microsoft designer to generate images with AI (text to image), and i like the results but there is no personal touch.
so, is there a way to feed an AI premade assets ? ( premade drawings or premade photos or premade 3d models), so that the result is more customized / personal art.

suggestions ?

thanks,

Yes ... these modern AI image generated often allow to simply drop an existing image of them and eg request to "create in the same style" or to "reimagine that picture"

Some identify if you give them copyrighted stuff - and propose to just create "similar" styles ...


bye
Ron

i was more talking that each 'thing' is inspired by another premade image which illustrates this thing.

so that the things and the graphics style of all generated images are consistent and can be used to illustrate a written story (of a book)

As there are models which are trained to create the same "figurine" each time (AI personalities) I guess it should be possible - but am not sure if that works for a "session" based thing (upload a picture and all the following things are trying to fit the style 100%). A special training could rule out "other" styles during training already and thus create a more "constant" output style.

Similar to the first incarnations in which you could have a picture reimaginated in specific styles ("van gogh") ... there they trained the models for special subsets.


Anyways ... I guess it should be possible, maybe it is already done - else it will come sooner or later. It requires a better "memory" of the "AI brain" so it does not pollute itself during the tasks you give to it. The more "abstract" the generated images are, the easier it becomes to have a consistent style. The more "characters" are on the images, the more likely you will spot differences.
I would for now use it more for a kind of "pre-viz" of specific sceneries ("Adventurers in a jungle, old temple ruins barely visible in the background...") but not for complete images in a comic strip.

if it was just a "10 images comic strip" (or less) I think you could create an AI model trained on short strips ... and let it generate a simple story (similar to the movie-clip-generating models ... which are short for now and do not tell "lengthier" stories).



bye
Ron

In some sad way I strongly feel you dehumanize yourself and your art in this way.
Embrace the effort, the struggle, the sharpening of your skill.

not really.

if we could define different areas / layers (one area / layer for each thing), and a personalized style of drawing, and let an ai mix everything, it would be a unique illustration with a personalized style, but done quickly.

however, i have not found that yet !