upload V1.0










README
If you appreciate this artistic style, please support the artist. Outstanding works are born from the artist's exquisite skill and creativity.
推荐的权重:0.95,推荐的模型NOOBAI-XL-Eps1.1
Recommended weight: 0.95, recommended model NOOBAI-XL-Eps1.1.
Trigger words:ibqy
以NOOBAI-XL-Eps1.1作为底模,使用LoKR算法训练,数据集的张数为:66。
观测到的缺点:
▶人物肢体很容易崩坏扭曲,猜测是训练集中人物张数过少。
▶在少量提示词的测试中,表现得并不好,须要丰富的提示词支持。
▶人物的脸很容易处于阴影里,推测是训练集处理时,对图片没有很好地标注。
作者推荐的:
▶你可以使用wide_shot,very_wide_shot进行构图,表现不错。
▶使用tag推荐:outdoor, scenery, sky, cloud, day, grass, dappled sunlight, sunlight, sunset, gradient sky, overexposure,
▶使用本Lora可以在负面提示词加入:(shade:1.1),(shaded_face:1.05),这样有助于提高画面表现。
▶质量词推荐:(very awa,masterpiece,best quality,highres,absurdres),ibqy,
Using NOOBAI-XL-Eps1.1 as the base model, trained with the LoKR algorithm, the dataset consists of 66 images.
▶The limbs of characters are prone to distortion and deformation, likely due to the small number of character images in the training set.
▶In tests with minimal prompt words, the performance was not satisfactory, requiring more detailed prompt support.
▶Characters' faces are often in shadow, possibly due to inadequate labeling of images during the training set processing.
▶You can use wide_shot and very_wide_shot for composition, which perform well.
▶Recommended tags: outdoor, scenery, sky, cloud, day, grass, dappled sunlight, sunlight, sunset, gradient sky, overexposure.
▶When using this Lora, it is recommended to add (shade:1.1) and (shaded_face:1.05) to the negative prompts to enhance the visual performance.
Quality words recommendation: (very awa, masterpiece, best quality, highres, absurdres), ibqy.
Under no circumstances is this LoRa permitted for commercial use, including but not limited to the sale of image packs or any similar activities.
upload V1.0
1. The rights to reposted models belong to original creators.
2. Original creators should contact SeaArt.AI staff through official channels to claim their models. We are committed to protecting every creator's rights. Click to Claim
