


Provide Checkpoint/three types of LoRA/reference-based character base prompt/theme. With these inputs, two images with the same composition are output: one anime-style and one realistic. You can use today’s theme or set your own for fun. Male-oriented themes are supported to some extent. If LoRA, etc. has trigger words, adding annotations after the base prompt is effective. LLM will process important instructions written as annotations.
Note: After conversion, the prompt has been adjusted to make exaggerated anime proportions as natural as possible. However, there are limits to correcting overly deformed proportions.
2026/6/11: During Flux2 image generation, part of the original prompt is now organized by LLM and used as a reference. Prompts are tuned to more clearly recognize character features and the target composition, and to fill missing parts using references.