What model was used for the anime results?

[vedantroy]

For this result:

What custom model was used?

[williamyang1991]

https://civitai.com/models/6526/studio-ghibli-style-lora

[vedantroy]

Thanks for the quick response -- how did you manage to load this LoRA into your method?
Did you use the method mentioned here: #39 (comment)?

More generally, do you mind posting the controlnet weight + prompts you used for this example?

I am trying to replicate this result as a starting point, but having a difficult time doing so.
I understand you are probably busy, so no worries if this isn't possible at the moment.

[williamyang1991]

Did you use the method mentioned here: #39

Yes

the controlnet weight + prompts you used for this example?

base model: Counterfeit
lora model: Ghibli
prompt: "ghibli style, a handsome man"
contronet strength: 0.7
denoising strength: 0.25

[zoeouyang2543]

hello：
For this result:
1、a cartoon tiger
2、a traditional mountain in chinese ink wash painting
3、Hermione Granger
What custom model was used?
do you mind posting the controlnet weight + prompts you used for this example?

[williamyang1991]

hello： For this result: 1、a cartoon tiger 2、a traditional mountain in chinese ink wash painting 3、Hermione Granger What custom model was used? do you mind posting the controlnet weight + prompts you used for this example?

Please refer to https://drive.google.com/file/d/1HkxG5eiLM_TQbbMZYOwjDbd5gWisOy4m/view?usp=sharing

[zoeouyang2543]

hello：
我把c站上的模型和lora转换成diffusers可以读的形式：
python ./diffusers-0.20.2/scripts/convert_original_stable_diffusion_to_diffusers.py --checkpoint_path ./Rerender_A_Video/models/counterfeitV30_v30.safetensors --dump_path ./Rerender_A_Video/models/Counterfeit --from_safetensors
python ./diffusers-0.20.2/scripts/convert_lora_safetensor_to_diffusers.py --base_model_path ./stable-diffusion-v1-5/ --checkpoint_path ./Rerender_A_Video/models/moxin.safetensors --dump_path ./Rerender_A_Video/models/moxin
然后在用这个方案进行加载 python ./Rerender_A_Video/rerender.py --cfg ./Rerender_A_Video/config/real2sculpture.json，
得到的结果不是项目中给出来的效果，难道只能用weiui才能复现你给的效果吗？麻烦解答一下
real2sculpture.json的内容是：
{
"input": "./Rerender_A_Video/videos/pexels_antoni_shkraba_8048492_540x960_25fps.mp4",
"output": "./Rerender_A_Video/videos/man/blend.mp4",
"work_dir": "./Rerender_A_Video/videos/man",
"key_subdir": "keys",
"sd_model": "./Rerender_A_Video/models/countfeit_ghibli",
"frame_count": 102,
"interval": 10,
"crop": [
0,
180,
0,
0
],
"prompt": "ghibli style, a handsome man",
"a_prompt": "RAW photo, subject, (high detailed skin:1.2), 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3",
"n_prompt": "(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, mutated hands and fingers:1.4), (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, disconnected limbs, mutation, mutated, ugly, disgusting, amputation",
"x0_strength": 0.95,
"control_type": "canny",
"canny_low": 50,
"canny_high": 100,
"control_strength": 0.7,
"seed": 0,
"warp_period": [
0,
0.1
],
"ada_period": [
0.8,
1
],
"freeu_args": [
1.1,
1.2,
1.0,
0.2
]
}

[williamyang1991]

我印象里用的是比v30更低的版本

[zoeouyang2543]

我印象里用的是比v30更低的版本

好的，谢谢，我试试counterfeitV30_v30.safetensors 更低的版本看看。
还有一个问题，config文件有什么特别注意的吗？您这边用的是哪个config文件执行的呢

[williamyang1991]

"a_prompt": "RAW photo, subject, (high detailed skin:1.2), 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3",
"n_prompt": "(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, mutated hands and fingers:1.4), (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, disconnected limbs, mutation, mutated, ugly, disgusting, amputation",
是用于realistic vision的
卡通风格用
"a_prompt": "best quality, extremely detailed",
"n_prompt": "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",

甚至extremely detailed可以去掉

[zoeouyang2543]

"a_prompt": "RAW photo, subject, (high detailed skin:1.2), 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3", "n_prompt": "(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, mutated hands and fingers:1.4), (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, disconnected limbs, mutation, mutated, ugly, disgusting, amputation", 是用于realistic vision的 卡通风格用 "a_prompt": "best quality, extremely detailed", "n_prompt": "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",

甚至extremely detailed可以去掉

您好，按照你的指引，得到的结果还是很奇怪，我下载的是v2.5的版本
会不会是某些参数需要配置呢，还是说合并模型和lora后调用，跟直接放在webui上使用差异还是很大的？
{
"input": "/home/notebook/data/group/oyp/Rerender_A_Video/videos/pexels_antoni_shkraba_8048492_540x960_25fps.mp4",
"output": "/home/notebook/data/group/oyp/Rerender_A_Video/videos/man/blend.mp4",
"work_dir": "/home/notebook/data/group/oyp/Rerender_A_Video/videos/man",
"key_subdir": "keys",
"frame_count": 102,
"interval": 10,
"sd_model": "/home/notebook/data/group/oyp/Rerender_A_Video/models/countfeit_25_ghibli",
"prompt": "a handsome man in cartoon style",
"a_prompt": "best quality, extremely detailed",
"n_prompt": "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",
"x0_strength": 0.75,
"control_type": "canny",
"canny_low": 50,
"canny_high": 100,
"control_strength": 0.7,
"seed": 0,
"warp_period": [
0,
0.1
],
"ada_period": [
1,
1
]
}

[williamyang1991]

你的prompt也和我的不一样
我的是 ghibli style, a handsome man

我不太记得我当时用的canny还是HED，你可以都试一下

[zoeouyang2543]

你的prompt也和我的不一样 我的是 ghibli style, a handsome man

我不太记得我当时用的canny还是HED，你可以都试一下

我都试过了，但是都得到的结果都不是你这边给到的examples，哈哈哈哈
会不会是某些参数的设置不一样呢，当时的json文件还有吗

[williamyang1991]

因为我离职回国了，服务器上很多文件就删掉了。
而且代码是中稿之后才整理的，整理前的代码还没有用json文件来指定标注，我基本都是用txt来记录实验设置的。
我翻了一下当时就记录了：
ghibli style, a handsome man, 0.7, 0.25, Counterfeit(Ghibli)
其中0.25对应1-"x0_strength"，0.7对应"control_strength"

[zoeouyang2543]

因为我离职回国了，服务器上很多文件就删掉了。 而且代码是中稿之后才整理的，整理前的代码还没有用json文件来指定标注，我基本都是用txt来记录实验设置的。 我翻了一下当时就记录了： ghibli style, a handsome man, 0.7, 0.25, Counterfeit(Ghibli) 其中0.25对应1-"x0_strength"，0.7对应"control_strength"

好的，谢谢您，祝好