I made an AI Comic using Stable Diffusion with Consistent Characters

I made this AI comic using Stable Diffusion, mainly using ComfyUI to generate the ai art images. I then created an animated story for the comic. A lot of photoshop was used to pull everything together. But all the images of people, backgrounds objects, were created with AI. #stablediffusion #comfyui #aiart Link to download my workflow: https://civitai.com/articles/5093 This video was originally a lot longer, but I cut out several mins of me explaining loads of the extra details because it was getting too long and boring. And this video is more of a showcase of what I made and how I made it, rather than a tutorial. But for all of those who are curious about the extra details I'll go over them here in the description for you: It is best to use simple clothes. Anything with details will end up being completely different in each image. And using the same person in the same clothes for the input also helps a lot. If I take off the img2img for my character, and just used controlNet for the pose, I can get something consistent, but the images it becomes waaaay to sexy for youtube. Like literally no trousers, on or thongs showing hahaha. Now, even if the character is consistent, it can take many tries to get the pose right, Or you might get the right pose, but the hands are messed up. Or the face is looking the wrong way. Etc, etc. Just have to keep spinning that wheel and tweaking till you land on something workable. It is pretty bad at doing expressions too. Like a toddler it only exists in extremes, happy, angry etc. But anything subtle, like confusion or disgusted or bored it just flat out ignores you. Therefore I just tried to lower or amplify weights on similar expressions to get something close. A lot of trial and error here too. For the objects, like the cup and table. Trick for this. Do it separately. Find a 3D cup online, take a screenshot, then do img2img. Now we have the same cup in each shot. Left to the AI, half the time the cups would end up as random junk. So the background for the coffee shop is actually only just 3 images. We can just use them over and over and move it depending on who's talking. Oooohh magic… and no one is the wiser. And one of the toilet shots, where I look in the mirror was created using Dalle-3 instead of stable diffusion. But I had to edit that one a lot in photoshop to get my reflection and the toilet reflection into it. If you have any other questions let me know. Hope you enjoyed it :) -------------------- -------------------- SOCIAL: Shout your random internet thoughts at me here: https://twitter.com/EthanDavd Stalk me and find out what I had for dinner 4 years ago here: https://www.instagram.com/ethandavd