Newsletter callbacks and images

Today, I was listening to a podcast and noticed how the speaker easily made callbacks to previous episodes. "If you want to learn more about x then check out episode y. I started working on a system that I can use with my newsletter to find related posts.

To start I’m focusing on creating indexes, similar to how I have weekly summaries. That way I can refer to individual posts, weekly themes, or other subject-related themes.

I’ve also been working on a new set of header images on the posts. Although I have about 5k images already generated, The newer models I’ve downloaded are much improved. I created what’s called an “aesthetic embedding” using 256 selected images from the previous batch.

This setup lets me generate similar images but with much better quality.

I might also try a batch-processing job where I use the original images as inputs. Since the prompts are saved in the file metadata I can generate new images based on the existing image parameters, but with the new model and aesthetic gradient. It should produce a similar image of better quality.