This is merged checkpoint Pony Diffusion V6 with others checkpoints.
Recommended Settings:
Sampler: DPM++ SDE / Euler a
Sampling Steps: 20 - 50
CFG Scale: 5
Positive prompt: score_9, score_8_up, score_7_up,
Negative prompt: score_6, score_5, score_4
TIPS: This model has strict token limits. Exceeding these limits makes it harder to follow prompts. Aim for about 100 tokens or fewer. To check your prompt's token count, use the token visualizer on the Run.app website. Note that symbols like "_", "-", "()", "[]", and even commas count as separate tokens.
To maximize efficiency:
Use commas only when necessary to separate categories.
Group similar descriptors together without commas. For example: long hair black hair ponytail hime cut, blue eyes
This approach reduces token count and helps the AI process related concepts together. It's more effective because the AI tries to make sense of words in each token group.
But hey its just a theory. But I hope this help though =-=
