So, was the result completely unexpected, or are you able to know in advance if an image have the potential to generate something interesting?
I knew that the Paulstretch effect in Audacity has the potential to produce these sorta psychedelic patterns when you give it these images in "planar RGB" format. Planar RGB is very simple. You take the R, G and B components of a picture, then you just have these 3 planes of R, G and B stored right next to each other that then get combined to produce a full-color image.
So then I wanted to try this idea in other color formats, just out of curiosity. There's this thing mostly in video called YUV/YCbCr. What that does is store the grayscale image (Y), the blue projection (U/Cb), and the red projection (V/Cr). It depends on the exact format, but I went with planar YUV 4:4:4, which just stores the Y, U and V components as planes of grayscale, blue, and red
right next to each other in some way that I'm not aware of without dropping the resolution of any component.
I went into FFmpeg, banged my head against the wall trying to get it to work, got it to convert that nice picture into YUV, put that into Audacity and Paulstretched it. Then I banged my head against the wall yet again trying to get FFmpeg to accept the nonsense it's being fed, and finally got that result. At least it does produce nice color patterns, so that time I spent wasn't completely in vain. Could've gone completely sideways; I'm very sure this technique hasn't ever been published as a databending method. I am completely on my own for the first and last thirds (the parts involving FFmpeg) of this process.
I think I should write about this. I don't see myself as an artist, I see myself as a scientist who just happens to produce art on occasion, so I do want to just publish this method as an extension of the general method as detailed
here and
here. Maybe I'll try with different color formats, since with FFmpeg, I get to mess with a wide range of those. I don't care that my work is entirely reproducible on any computer with the right know-how. That's the point.
Edit: I have no clue how yuv444p, the pixel format I used is supposed to work. It's not 3 planes of Y, U and V stacked together consecutively.
Did you choose the original picture because of the rainbow?
It's pretty much that. I knew that the whole YUV thing basically splits images into the brightness component and two color components, so I wanted a colorful image so that the effect has something to work with. Effects cannot create new data, they can only manipulate it.