Processing notes :
Hello,
I made an upgrade to stablediffusion model, released these last days. More free, available under colab, but also on some websites.
- Now a couple of sites are heavily relaying the news coming from this AI image synthesis scene (https://multimodal.art/news). It will become most probably mainstream and disruptive in the coming weeks.
- The next step, beyond the worldwide availability to draw by instructing computers in natural language, will be animation and movie, with help of Ai, from idea to production.
- Prompt engineering is already a new job for people skilled in writing. Visual creativity will explode in mixing all already existing graphic styles included in the initial dataset of the AI synthesis image model.
- A lot of people amongst digital artist are already fully productive in that way and @kaliyuga is becoming really famous on the scene as well as @lavista for their active studies and grounded participation by releasing their own fine-tunes models.
Think about digital sampling in music and its use nowadays.
- Legal debate will be rage but at the end, this tech will be used.
- Censorship and biases also will be policed.
This my pure feeling of the trend i observe at the moment, nothing really scientific here.
About my upgrade...
It took me a couple of hours to prompt this new model with some satisfaction. All these new pages were raw outputs from a prompt, without any post-processing or re-iteration via init images. The parameter allows to play images derivation to keep more graphic coherence. Anyway, at this stage, if graphical coherence is increased, the style is still waving.
Character design is still a challenge but there are strategies in prompting that come closer to satisfying results : naming a known popular fiction or real person is a usual way.
I'm trying to keep a more descriptive symbolic way to prevent using already existing charachter. Wording derivation of the verse _'A blue hat dressed lady' _was enough to get this character of Julia along the pages, with all models, but with more expression and quality in stablediffusion.
Action scene are still under study. It's probably something other master already in their prompting but i need deeper investigation on that side.
My next steps is now investigating deeper comics scenarization and visuals effects, with help of Ai and some real books.
The Imaginarium graphic novel is probably reaching an open-ending episode. Julia is now taking knowledge of her own nature and it opens a lot of other quests.
I recognized the scenario is difficult to follow, quite mixed with self-esoteric cultural and technical references, moreover unleashed in some troubled graphic styles. But that were the constraint of the experiment. My total amateurish approach will most probably be now taken in hands by professionnal to produce really stunning graphic novel artworks. But anyway, i will probably stil produce my own delirium, to keep my little alien living inside me having fun. In music i like to play and try very different styles, so it appears i want to continue this in my own metaverse or imagination, whatever you named it.
i'm also thinking to put my creation in pdf into nftshowroom...
The issue in my cognitive case is the flow of productive ideas
- a volume decyphering all my inspirational storytelling references
- a volume explaining in comics how to use AI tools image synthesis
- a volume of new side stories linked to Imaginarium
- a volume of classic novels illustrated by AI in comic style
- approaching visual situasionist propaganda
- illustrating underdog amateur board games
- approaching comics animation
- mixing deeper gpt-3 and Ai drawing etc.
But ok, i never follow any list ;-)
My previous posts are here.
Now also some periodical chirping news via https://twitter.com/Dbddv01
My Greetings to all the Alien Art Community for their encouragement.