Try size: 768 x 1024 or 768 x 1280

Weights: 0.7~1.0

Clip skip: 2

Sampling steps: 10~20 (ZIT) / 30 or above (ILXL)

Training material from Jimeng and Banana and Z-Image Turbo

Note: You can use the short danbooru format because this lora trained with WD14 tags; Alternatively, you can provide the full context using gemini, chatgpt, or grok, because this lora trained with natural language captions. The proportions of a figure are usually related to the furniture, so you can describe the size of the bed and chair to fine-tune the proportions of the figure. Alternatively, you could add "from above" and "wide shot" to make the girls on the first floor appear smaller. / 你可以使用簡短的danbooru格式，因為這lora有經過wd14的tags訓練；也可以交給gemini、chatgpt或grok提供完整的上下文，因為這lora有經過natural language的caption訓練。人物的比例大小通常跟傢俱有關，所以可以描述床跟椅子的大小，來微調人物的比例。或者加上 "from above" 和 "wide shot"，讓一樓的女孩們比例變小。(ZIDT)

Note: If the face is blurry due to distance, it is recommended to enable "ADetailer" and draw the face with "Inpaint denoising strength: 0.6". You can replace "kawaii theme" with any color, such as "black theme", "white theme", etc. If the girl is difficult to appear or too far away, please add more details about this girl, or adding the prompts "upper body, close-up, portrait," and placed at the beginning of the prompt. / 如果因為距離太遠導致臉部模糊不清楚時，建議開啟"ADetailer"以"Inpaint denoising strength: 0.6" 繪製臉部。可以將"kawaii theme"替換成任何顏色，例如"black theme"、"white theme"等等。假如女孩很難出現或太遠，請添加更多關於這個女孩的描述，或距離提詞"upper body, close-up, portrait,"，並且放在提詞的首位。(ILXL)

Search AI Models & Addons

very few results