I generated a ton of similar ai images sorted them by difference and generated a video using morphing effects in ffmpeg. It morphs between images for a second, then shows a static image for a second at 30 fps, and just does that, never repeating. It outputs a text script that ffmpeg can use and gives you the command to use:
I have a program that does a high speed diff between the images to generate the list, which looks like this:
reassembled_list.txt (Size: 11.81 KB / Downloads: 5)
The video is here, it has bs triggers embedded:
The fact that the flux 2 klein 9B model is so consistent helps make this effect look good. It takes me days to generate a few minutes of this (I do not have a GPU to speed this up) and I need to tune in the search width to get the best effects on the morphing. But I think I will be able to create some interesting videos eventually.
The plan it so create an hour long video, use it as the background video with other videos overlaid on it and combine it with audio story using another program that read text and embeds sound effects and background music. I will layer a high speed flow of text that is in one spot on screen that is like a direct download of information into your head, it is faster than your internal monolog can keep up with, but you sill comprehend the text fully. No inner monolog bypasses some conscious filters on what you are downloading into your brain.
This is using local ai generation of images, local ai generation of speech, ai helped create the image sorter and the story teller programmer and a high speed text presentation software and programs to split and merge audio and video streams.
This is all a new way of doing things. These are interesting new times we live in now.
(This post was last modified: 27 Feb 2026, 15:02 by Like Ra.)
Code:
ffmpeg -f concat -safe 0 -i reassembled_list.txt -vf "minterpolate=fps=30:mi_mode=mci:mc_mode=aobmc:me_mode=bidir:me=esa:search_param=64:vsbmc=1:scd=none" -pix_fmt yuv420p +faststart morph_slideshow3.mp4I have a program that does a high speed diff between the images to generate the list, which looks like this:
reassembled_list.txt (Size: 11.81 KB / Downloads: 5)
The video is here, it has bs triggers embedded:
The fact that the flux 2 klein 9B model is so consistent helps make this effect look good. It takes me days to generate a few minutes of this (I do not have a GPU to speed this up) and I need to tune in the search width to get the best effects on the morphing. But I think I will be able to create some interesting videos eventually.
The plan it so create an hour long video, use it as the background video with other videos overlaid on it and combine it with audio story using another program that read text and embeds sound effects and background music. I will layer a high speed flow of text that is in one spot on screen that is like a direct download of information into your head, it is faster than your internal monolog can keep up with, but you sill comprehend the text fully. No inner monolog bypasses some conscious filters on what you are downloading into your brain.
This is using local ai generation of images, local ai generation of speech, ai helped create the image sorter and the story teller programmer and a high speed text presentation software and programs to split and merge audio and video streams.
This is all a new way of doing things. These are interesting new times we live in now.
