add technical setting and special tag with minor fix
Browse files
README.md
CHANGED
@@ -12,7 +12,7 @@ tags:
|
|
12 |
A **merge** stable diffusion model specialize in **anime**, which improves anatomy of anime characters, especially with **eyes** and **hands**, without losing anime objects (like substances or charaters).
|
13 |
much better for working at 0.9K-1.2K resoultion, or use Hires.fix instead.
|
14 |
Provide in mk1 and mk2 version. Personally mk1 works better, but mk2 give out more vivid pictures.
|
15 |
-
融合炉2.5D
|
16 |
其在900-1200像素的分辨率下工作较好,或者可以使用高清修复改善其高分辨率表现。
|
17 |
提供mk1和mk2两个版本。个人感觉mk1版工作的更好,但是mk2版本能生成更生动的图像。
|
18 |
|
@@ -36,8 +36,10 @@ Provide in mk1 and mk2 version. Personally mk1 works better, but mk2 give out mo
|
|
36 |
|
37 |
# Usage(使用方法)
|
38 |
|
39 |
-
Use as normal stable diffusion model package v1.x, no external yaml config is needed.
|
40 |
-
|
|
|
|
|
41 |
|
42 |
|
43 |
# Tags(描述词)
|
@@ -47,7 +49,10 @@ Positives as you like, maybe less quality words works better. You can get inspir
|
|
47 |
**Negatives Example:** *(bad_prompt), cleavage, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artists name*
|
48 |
正面填写你喜欢的描述词,也许更少的质量描述词能使其工作的更好。你可以在上面的预览图描述词中得到灵感。
|
49 |
**负面描述词最好用基本负面,或者简单的把它们替换成bad_prompt这个嵌入模型。**
|
50 |
-
|
|
|
|
|
|
|
51 |
|
52 |
|
53 |
# Bias(不足)
|
@@ -65,11 +70,11 @@ Sometimes long necks appear. Still hazy a bit. Under some theme will produce wro
|
|
65 |
**Round1** animefull-latest(NovelAI)+64in1(Private, from a Chinese AI community NovelAI 中文频道) sum rate0.4
|
66 |
**Round2** ()+AbyssOrangemix2_nsfw(WarriorMama777) sum rate0.2
|
67 |
After baked in vae-ft-mse-840000-ema-pruned(StabilityAI) VAE, pruned ema, compressed to FP16, get MergeStove2.5D_mk1.
|
68 |
-
**第一轮** animefull-latest(NovelAI)+64in1(
|
69 |
**第二轮** ()+AbyssOrangemix2_nsfw(WarriorMama777) 加权和模式 比率0.2
|
70 |
嵌入vae-ft-mse-840000-ema-pruned(StabilityAI)这个VAE模型后,去掉EMA权重,压缩为FP16格式,得到MergeStove2.5D_mk1模型。
|
71 |
|
72 |
-
**Round3** MergeStove2.5D_mk1+Anmokomergetest1(Private, from a Chinese AI community NovelAI
|
73 |
After baked in vae-ft-mse-840000-ema-pruned(StabilityAI) VAE, pruned ema, compressed to FP16, get MergeStove2.5D_mk2.
|
74 |
-
**第三轮** MergeStove2.5D_mk1+Anmokomergetest1(
|
75 |
嵌入vae-ft-mse-840000-ema-pruned(StabilityAI)这个VAE模型后,去掉EMA权重,压缩为FP16格式,得到MergeStove2.5D_mk2模型。
|
|
|
12 |
A **merge** stable diffusion model specialize in **anime**, which improves anatomy of anime characters, especially with **eyes** and **hands**, without losing anime objects (like substances or charaters).
|
13 |
much better for working at 0.9K-1.2K resoultion, or use Hires.fix instead.
|
14 |
Provide in mk1 and mk2 version. Personally mk1 works better, but mk2 give out more vivid pictures.
|
15 |
+
融合炉2.5D是一个**动漫风格特化**的稳定扩散模型,由**多个模型融合**而来,专门改善动漫人物的身体结构,特别是**眼睛**和**手**,同时不会丢失任何动漫中的对象(物体、人物等)。
|
16 |
其在900-1200像素的分辨率下工作较好,或者可以使用高清修复改善其高分辨率表现。
|
17 |
提供mk1和mk2两个版本。个人感觉mk1版工作的更好,但是mk2版本能生成更生动的图像。
|
18 |
|
|
|
36 |
|
37 |
# Usage(使用方法)
|
38 |
|
39 |
+
Use as normal stable diffusion model package v1.x, no external yaml config is needed.
|
40 |
+
**Recommand settings: Steps: 15, Sampler: DPM++ SDE Karras, CFG scale: 7-9, Denoising strength: 0.7, Hires upscale: 2, Hires upscaler: Latent**
|
41 |
+
用作正常的稳定扩散模型包v1.x,无需额外的YAML配置文件。
|
42 |
+
**推荐设置:采样步数:15,采样器:DPM++ SDE Karras,提示词相关性:7-9,去噪强度:0.7,高清修复放大倍率:2,高清修复放大器:Latent**
|
43 |
|
44 |
|
45 |
# Tags(描述词)
|
|
|
49 |
**Negatives Example:** *(bad_prompt), cleavage, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artists name*
|
50 |
正面填写你喜欢的描述词,也许更少的质量描述词能使其工作的更好。你可以在上面的预览图描述词中得到灵感。
|
51 |
**负面描述词最好用基本负面,或者简单的把它们替换成bad_prompt这个嵌入模型。**
|
52 |
+
**负面描述词示例:** *(bad_prompt), cleavage, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artists name*
|
53 |
+
|
54 |
+
**Use "blue eyes black middle" description can get huge improvement on pupil at low resolution! Colors can change as your preferance.**
|
55 |
+
**使用"blue eyes black middle"这样子的描述词可在低分辨率下极大的改善对瞳孔的描绘!颜色可以改为你喜欢的。**
|
56 |
|
57 |
|
58 |
# Bias(不足)
|
|
|
70 |
**Round1** animefull-latest(NovelAI)+64in1(Private, from a Chinese AI community NovelAI 中文频道) sum rate0.4
|
71 |
**Round2** ()+AbyssOrangemix2_nsfw(WarriorMama777) sum rate0.2
|
72 |
After baked in vae-ft-mse-840000-ema-pruned(StabilityAI) VAE, pruned ema, compressed to FP16, get MergeStove2.5D_mk1.
|
73 |
+
**第一轮** animefull-latest(NovelAI)+64in1(私有,来自中国AI社区NovelAI 中文频道) 加权和模式 比率0.4
|
74 |
**第二轮** ()+AbyssOrangemix2_nsfw(WarriorMama777) 加权和模式 比率0.2
|
75 |
嵌入vae-ft-mse-840000-ema-pruned(StabilityAI)这个VAE模型后,去掉EMA权重,压缩为FP16格式,得到MergeStove2.5D_mk1模型。
|
76 |
|
77 |
+
**Round3** MergeStove2.5D_mk1+Anmokomergetest1(Private, from a Chinese AI community NovelAI 中文频道, Download [Anmokomergetest1](https://huggingface.co/LieDeath/Anmokomergetest1).) sum rate0.4
|
78 |
After baked in vae-ft-mse-840000-ema-pruned(StabilityAI) VAE, pruned ema, compressed to FP16, get MergeStove2.5D_mk2.
|
79 |
+
**第三轮** MergeStove2.5D_mk1+Anmokomergetest1(私有,来自中国AI社区NovelAI 中文频道,下载[Anmokomergetest1](https://huggingface.co/LieDeath/Anmokomergetest1)。) 加权和模式 比率0.4
|
80 |
嵌入vae-ft-mse-840000-ema-pruned(StabilityAI)这个VAE模型后,去掉EMA权重,压缩为FP16格式,得到MergeStove2.5D_mk2模型。
|