Pony
Pony Diffusion: a line of models trained further from SDXL, with its own quality and rating tags.
What it is
Pony Diffusion V6 XL is an SDXL model retrained at large scale on stylised and anime material. Many further fine-tunes are built on it. Technically it is SDXL: it loads, runs and sizes exactly like SDXL. What differs is how it reads the prompt.
Files and nodes
One checkpoint in models/checkpoints, Load Checkpoint, Empty Latent Image, as for SDXL. Many Pony models want CLIP Set Last Layer at -2; the model page says.
Settings that work
SDXL sizes (1024×1024, 832×1216); 25 to 30 steps; cfg about 5 to 7; euler_ancestral or dpmpp_2m. Prompt: tags, starting with its score tags (score_9, score_8_up, score_7_up, ...) as the model page lists, plus source and rating tags the page documents.
What fits
LoRAs trained for Pony fit best. SDXL LoRAs load but often behave differently; Illustrious LoRAs are hit and miss. SD 1.5 and Flux add-ons do not work.
How the app helps
Shown as Pony on model cards, separate from plain SDXL, so a LoRA's card can tell Pony LoRAs from SDXL ones. The lesson Pony and Illustrious: tags instead of sentences writes a tag prompt.