HPL in the 18th century. Would he have been a College Hill gent as he hoped, or (more likely) a jobbing waterfront printer and local news-sheet editor?

No, not the result of the new Flux Kontext, which is a bit ‘off’ in terms of this sort of thing since it works from a single image. Also slow. Instead I made an SDXL LoRa of his face and shoulders, and used it with the superfast four-step SDXL checkpoint realvisxlV50_v50LightningBakedvae.safetensors   Considering this generates a 1024px image in about 3 seconds on a very modest 12Gb graphics card, the quality here is acceptable. Not superb, a little glitchy if you look hard for it… but acceptable. And yes, tweaking the prompt will cause him to smile. But I thought I’d spare the reader that.

The new Flux Kontext does have its uses though, including easy manga-fication of images.