After making indy-news.streamlit.app and its counterpart GPT called Indy News Assistant. I wanted to make a comicbook series that would be a fun way to get that same news. So I set out to build it and thought "and they will come" ;)

The very first episode came about after some tinkering with AI agents, some scripting, and a lot of manually prompting and Dall-E 3 for images. At first OpenAI wouldn't let me get away with my truth (independent news) prompts to create the artwork. It just told me over and over about violating their policies etc, and I got disillusioned and turned to Stable Diffusion. Which was waaaay to much of a hassle to get to the right combo of base model and comicbook LoRa. But After trying again some days later I found that Dall-E 3 will cooperate if prompted via the api, instead of manually via chatgpt, YAY! Energy started flowing again and I continued on the AI agency and the automation and decided I will make it work even if I have to apply some human intervention. (That part really blocked me as I am an automation engineer, and I just had to go for results instead of eternal architecture mongering ;)

And the results are quite fascinating I think. The combo of beautiful imagery with subtext kinda gets the news across, but for now it seems more of a wacky experiment that can please the senses. I do hope the future will bring better prompt following image makers, but for now I just have to slap what I have into shape sometimes ;p

The process for the first episode went more or less like this: I first asked my indy-news api (as in use by indy-news.streamlit.app/Youtube) for last week's news from channels aljazeeraenglish,democracynow,thegrayzone7996,glenngreenwald and saved it as input.json. This was then forwarded to my AI comic agency (consisting of a researcher, writer, illustrator and editor), and the researcher came up with input.json, which was handed over to the writer and editor, that came up with the final comicbook storyboard, resulting in comic.json. I then gave it to the artwork workflow to iterate over the scenes involving illustrator to create the artwork for each scene. By now I have it all automated and I just give it the date/range I want to make the comic for, et voila! (I do have to tinker with the monotonous style and camera perspective, which is prompt-tuning in progress ;)