AI still hates us. Ai remember waterworks.

More about AI still hates us. Ai remember waterworks.

Well, figuring out how to get the red piping back is the only challenge, everything else I can get to work. I guess it keeps me off the streets. Or it would, if it didn't take a bit over 7 hours for each attempt. I'm only trying 5 second clips until I can figure out how to get the red stripes above the hemline. I already have a 10 second clip I'm perfectly happy with except for that.

I have another option I've already tried once, with not much luck. Take the last frame I have where the skirt is right, and move her over to approximately the same location she's standing now, and make a jump-cut in my compilation video. Problem with that idea is, again, continuity. Matching the blazer and tie lying on the sideboard dresser is just as big a challenge as getting the skirt to match what it looked like until the camera zoomed out too far. Never expected that to be as big an issue as it is turning out to be.

I suspect a large part of the problem is having no idea what model ProVideo uses. Which says any prompting I figure out for single images is worthless. I've already determined that prompts which produce the correct image for Flux do not produce the same results on Photonic. I can't just edit the red piping back into the last frame of the previous video, because the shot is framed too tightly for it to be visible. I suppose I can have a long talk with ChatGPT or Grok and MAYBE they can dig up the info on whatever the heck ProVideo is looking for. I already had one session trying to get a prompt figured out that way, but I may have to dig a bit deeper.

It's not just DDG's model, though, whatever it really is. I've been experimenting a bit with my Stable Diffusion setup at home, and was rather surprised how weird Epic is about trying to correct that image, too. I thought that would be fairly simple inpainting to set up for Replacer, but no such luck. 62 different ways of writing that prompt, and I've yet to find a way to get anything close to what I want. Even feeding it prompts that Clip Interrogator gave me describing the image from before DDG decided to forget it was there, or that ChatGPT or Grok claimed would work with that model. I've pretty much decided AI doesn't stand for 'artificial intelligence', it stands for 'annoying idiocy'.