T2I 스탭수 변경 후 CamelliaMix 조언을 구합니다

parameters	(full body, from_front:1.3), (1 African alchemist girl from Ghana, tanned_skin:1.25), (hands on hips:1.2), solo, (gigantic breasts + muscular:1.4), (best black cape of Ghana, underboob:1.35), (detailed witch_hat and shoes), whole_body, finely detailed, navel, (smile:0.8), Ghana, (white background:1.3), <lora:film:0.8> NAI content_copy WEBUI content_copy Original content_copy
negative_prompt	nipples, (from above, from below, from side:1.5), nsfw, paintings, sketches, (worst quality, low quality, normal quality:1.8), lowres, ((monochrome)), ((grayscale)), bad anatomy, Deep Negative, (fat:1.2), facing away, looking away, tilted head, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped hands, username, blurry, bad feet, cropped, poorly drawn hands, poorly drawn face, mutation, deformed, jpeg artifacts, signature, watermark, extra fingers, fewer digits, extra limbs, extra arms, extra legs, malformed limbs, fused fingers, too many fingers, long neck, cross-eyed, mutated hands, bad body, bad proportions, missing fingers, missing arms, missing legs, extra digit, badhandv4, EasyNegative, NG_DeepNegative_V1_75T, (multiple view, reference sheet, different eyes, small breasts, medium breasts, flat chest, animal ears, elf ears, cat ears), animal, (indoor, outdoor, black background, grey background:1.3), unclear face, kneeling, strabismus, (shadow on ground), (teeth, effects, short legs:1.3), (sandals), nail art, nail, (head cut, shoes cut, clothes cut) NAI content_copy WEBUI content_copy Original content_copy
steps	30 content_copy
sampler	DPM++ SDE Karras content_copy
cfg_scale	8 content_copy
seed	2121635767 content_copy
size	512x768 content_copy
model_hash	2eb0c2a23a content_copy PROMPT SEARCH: CamelliaMix open_in_new
clip_skip	2 content_copy
denoising_strength	0.4 content_copy
Copy All	NAI content_copy WEBUI content_copy Original content_copy

항상 많이 배우고 있습니다.

현상) 패치 후 위의 이미지 처럼, 강조를 해도, black 머시기나 dark skin 등을 입력해도 이전과 같은 수준의 인종 구분을 못하고 있습니다.

질문) 30스탭에 8~10 스케일 안에서 해결할 만한 태그 조합이 있을까요? 분명 데이터는 있으니 구현은 어떻게든 될 거 같은데 태그를 잘못 입력한 걸까요? 이젠 이전처럼 지역/국가를 입력하는 것만으로는 인종구분이 어려운 걸까요? 조언을 구해봅니다.