The first time I tried seed travel, I used a relatively simple prompt. After that, I wanted to try something a little more impressive. I gave Stable Diffusion 1.5 the prompt that I used to generate the turtle diorama (see my post Dioramas I) and had it generate a few more images:
From these, I narrowed it down to a set of 8, in the following order, to make a seed travel with. I didn't notice until it was too late, that the 7th image was also set to be the first image, so I actually ran a seed travel between 9 images, one of them used twice.
And here is the final seed travel animation:
But I didn't stop there. When I tried to do some other animations, I ran into some problems. So I switched from Automatic1111 to SD.Next, by Vladmandic, which I've been pretty happy with. I tried this process again, but using a version of Stable Diffusion 1.5 that had been further trained, called DreamShaper. Here are the images I generated to start with:
As you can see, the colors are a bit more vibrant and the images look more like digital paintings and less like oil paintings. From here, I narrowed it down to the following images in the following order:
And here is the final animation:
These illustrations and animations were created using Stable Diffusion 1.5 in Automatic1111 and DreamShaper 7 in SD.Next.
