From what I understand about AI video generators you can provide an initial frame and a ending frame and give a prompt for the in-between. Maybe I'm wrong. But is there a chance of making some AI sloppa turn comics/manga into AI videos? Like take each panel and feed that as start and end frames for a sequence, the script from the speech and thought bubbles and punch that into a robit to create an animated adaptation? What if it could turn the whole thing into an automated process? Like have some robit pull the panels, inset or not and interpret them as frames for the sequence. An LLM using OCR can read the dialogue and narration and perhaps create it all into prompts for a final rendering generative model to create the video from it all.Just a thought, it'll be total shit, but fuck it, lets see what it'd look like.
>>110051901yes now ask an AI and vibecode it
>>110051901>it'll be total shit, but fuck it, lets see what it'd look like.yes, anon. this is what the world desperately needs. more slop
People are making porn with voice acting and all from mainstream content on >>>/gif/ It's getting better and better.
>>110051901i think we're closer to this than you think-the current popular video generation model already wants a prompt format that's annoying enough to write by hand that it's common already to use tools to generate the shot by shot video script from reference images or just text ideas. this can easily be a page from a manga (picrel)-video generators today can already use image references and aren't just first frame/last frame. which means you can provide a script and a visual reference sheet for each character so there's no real need to generate start/end frames-OCR is already done by the current multimodal models of the day and isn't really a separate consideration anymorein practice local models have enough jank you'd still want to correct by hand ahead of time (in picrel it confuses the speakers in the very first shot), and you'd need good references for not just the character's live action appearance but also their voice so they don't change from scene to scene, but you can do what you're suggesting right now. and if you used seeddance instead of something local like minimax it would even probably look good
>>110055879actually for some reason i thought you'd mentioned live action somewhere in there, it's even easier for animation of course since reference images are readily available, but the process is the same.
the technology is almost ready for real production. Hollywood is a shambles right now.
>>110055879Post the outputs you do get, outsource your autism if you didn't get enough tylenol as a baby..
>>110051901You can do it right now, it's just most people are too lazy to even do that.
