did they use an LLM

It would be a bit strange to use a language model on an image. A diffusion model would be more appropriate. That said, I think Hammond’s explanation is most likely.

I don’t believe there is any interpolation between frames. The entire sequence consists of 22 images acquired over a roughly 20 minute period. I believe the melting effect you’re referring to could the loose regolith dropping into the fractures between the plates. In addition the original was an MP4 file (see link in the post). This Lemmy instance does not support MP4 uploads so I had to convert it to a GIF, there could be some image artifacts from that conversion process, so try watching the original MP4 animation in the link :) Or check the original HazCam images on the mission image server.

As Paul Hammond mentions, the “melting” is due to material sliding downhill. During the abrasion, which works partly by percussion, you can actually see a pebble sliding downhill (between this frame and this frame, which were taken only one minute apart), to the bottom right of the abrasion bit itself (near the centre of the image).

It’s not always apparent from the images, but the rover is currently on a fairly steep slope; we’re still parked on the exterior/outboard side of the Jezero Crater rim. I wonder how difficult it would be to make the 3D images I used to see from earlier rover missions…