Fooocus Advanced #117
Replies: 115 comments 270 replies
-
|
I could put an option so that the image is not seen at the same time that it is being processed, since it consumes too much time and memory, thanks and congratulations for this great work, you do greetings.. |
Beta Was this translation helpful? Give feedback.
-
|
Gracias saludos |
Beta Was this translation helpful? Give feedback.
-
|
Thank you, you are the best |
Beta Was this translation helpful? Give feedback.
-
|
when are you guys are adding control net and image to image? |
Beta Was this translation helpful? Give feedback.
-
|
and Fooocus step by step will remake to Automatic) |
Beta Was this translation helpful? Give feedback.
-
|
Can't seem to generate images of crying |
Beta Was this translation helpful? Give feedback.
-
|
Is it possible for foocus to generate a similar image based on an uploud image or an image from a link? |
Beta Was this translation helpful? Give feedback.
-
|
I want know this too |
Beta Was this translation helpful? Give feedback.
-
|
This would be amazing to be able to upload an image as the seed photo example |
Beta Was this translation helpful? Give feedback.
-
|
me too |
Beta Was this translation helpful? Give feedback.
-
|
I think the integration of controlnet could do this. |
Beta Was this translation helpful? Give feedback.
-
|
if anyone has any ideas or thoughts on this please DM me https://twitter.com/jay_azhang |
Beta Was this translation helpful? Give feedback.
-
|
If it's possible to make this request, I would really like to be able to use your UI with the 1.5 models. xD |
Beta Was this translation helpful? Give feedback.
-
|
Fooocus tới đây có controlnet không ad? nếu có thì kế hoạch là bao giờ có ? |
Beta Was this translation helpful? Give feedback.
-
|
You can, I've used DreamShaper 8 with it. It just doesn't look nearly as good as the SDXL models. If you use 1.5 models with it I recommend going to the Advanced Tab and turning "Sampling Sharpness" to 0. In any case, the 1.5 models don't look quite right in Fooocus, so I agree it would be awesome to have a SD 1.5 version for Fooocus. |
Beta Was this translation helpful? Give feedback.
-
|
sounds like "tile_colorfix+sharp" option in sd-webui-controlnet. anyway, nice work~ |
Beta Was this translation helpful? Give feedback.
-
|
哥们为你点赞！！牛鼻！！期待更加好用！ |
Beta Was this translation helpful? Give feedback.
-
|
Is it possible to add other based models from SD1.5? |
Beta Was this translation helpful? Give feedback.
-
|
Hi lllyasviel, Your "Fooocus" is a great job. I like it so much. Though I am eager to request you to develop all the features that Automatic 1111 has, but my prime request is, why don't you rewrite all the extensions of Automatic 1111. Since the comfort level with your work is far better, In A1111, starting from img3img, controlnet and others I personally come across lot of bugs in everyday. For example the embeddings tab fail to load at times. The preview of controlnet does not show up but use to function. Like this I can keep on saying many. Please consider developing your own extensions for A1111. About Fooocus, it is marvellous. I love working with more models and Loras.(Hope I can load other XL models and Loras). Looking forward to further improvements in Fooocus, that you may have in your action plan. Thank You for the efforts Good Luck to you |
Beta Was this translation helpful? Give feedback.
-
|
May be for me and you. For experts, it is a kinda compiling error clearing. However thank you for motivating him. |
Beta Was this translation helpful? Give feedback.
-
|
@tmprabubiz Support for img2img and embeddings is available in Fooocus-MRE variant of mine - it's dedicated for a bit more advanced users, who want to have a bit more control over the generation process. |
Beta Was this translation helpful? Give feedback.
-
|
I am using your variant only. I am seeing wonders in it. It is almost midjourney. Particularly the short prompting is a bliss. |
Beta Was this translation helpful? Give feedback.
-
|
@tmprabubiz My YT skills are pretty limited, but I provided examples illustrating how different MRE features work in MRE Wiki. If something is missing or unclear, please let me know - I will update the wiki, then. |
Beta Was this translation helpful? Give feedback.
-
|
Thank You so much |
Beta Was this translation helpful? Give feedback.
-
|
Thank you, lllyasviel (and camenduru) for the CoLab link. I was getting frustrated with my system's slowness. |
Beta Was this translation helpful? Give feedback.
-
For those following along, the webui does technically support the sharpness parameter. It's identical functionality to SAG, just with a different control scheme, and has an extension available. It's pretty easy to verify this by viewing the git blame (although the commit messages don't make this very clear). https://github.com/lllyasviel/Fooocus/blame/18a8c6655208c31216b83bd93381cb12993597c2/modules/patch.py#L351 https://github.com/ashen-sensored/sd_webui_SAG (README mentions need of a patch -- that is no longer needed) This of course doesn't support SDXL currently (I haven't tested but the slightly altered model architecture leads me to believe it wouldn't function), but that could always change. ComfyUI does not provide any sort of implementation or custom node for this currently as far as I'm aware. |
Beta Was this translation helpful? Give feedback.
-
|
Installed "sd_webui_SAG" in A1111. But I could not find it in the user interface. May you please explain the right way of installation and usage or mention if there is any updated tutorial available online. Thank You. |
Beta Was this translation helpful? Give feedback.
-
|
swapping base models is going to make this a killer app |
Beta Was this translation helpful? Give feedback.
-
|
生成不了图，谁能告诉我这个是什么情况：Error |
Beta Was this translation helpful? Give feedback.
-
|
issues92: |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
要下载两个模型，放Fooocus\models\checkpoints里。 |
Beta Was this translation helpful? Give feedback.
-
|
FOC作者使用采样器和精炼器都是精心设置过的，所以对提示词要求很低，随便写也能出好图，速度模式30步 质量模式60步，本身就慢，如果是显卡不好就更慢（实测3050 1024 1张图4分钟），你如果想要自己设置采样器加速 可能需要的是这个 Fooocus-MRE 魔改版，增加了一些功能~ |
Beta Was this translation helpful? Give feedback.
-
|
能出图不，一直在最开始出不了图 |
Beta Was this translation helpful? Give feedback.
-
|
Fooocus is amazing, I'm a heavy Midjourney user who has tried SD several times and can never get my head around the settings or the UI, and I'm quite nerdy. This is great, thank you. I wonder in future versions the style, seen in the advanced features, could be auto-selected based on phrases or words written in the prompt? |
Beta Was this translation helpful? Give feedback.
-
|
I wouldn't want my selected style to be affected by my prompt , I want it to remain how it is now and i can combine a lot with prompt and selected style. You can write anything in prompt as you want and your image will be based on that input no need to change style automatically |
Beta Was this translation helpful? Give feedback.
-
|
How can I change the port when I run fooocus on windows? |
Beta Was this translation helpful? Give feedback.
-
|
Add |
Beta Was this translation helpful? Give feedback.
-
|
Hello and thanks a lot for your hard work. Had installed it just after release but didn't take the time to play with it, now going back to it after a long and pretty frustrating session with Automatic1111, and damn does your approach feel good instead of endless finicky tinkering of often conflicting parameters. One thing though which I think is sorely lacking compared to Automatic1111 is the use of weights and the prompt editing feature. To me, it feels like that is an integral part of focusing on prompting. Any chance you might consider adding them in the future ? Thanks in advance ! |
Beta Was this translation helpful? Give feedback.
-
|
things like (high quality:1.1) are supported |
Beta Was this translation helpful? Give feedback.
-
|
Dear Lvmin could you please advise how to resolve issue?
|
Beta Was this translation helpful? Give feedback.
-
|
another question, is it possible i could make photos with 800x800, where to modify the script if could. really appreciate your support. its amazing AI software |
Beta Was this translation helpful? Give feedback.
-
|
i have the same problem，cpu 12400 100% but gpu 1660s 6G not work ，mem 16G error |
Beta Was this translation helpful? Give feedback.
-
|
I met this issue too, my memory config mets the requirement, however, I found my disk for the model stored was almost full... after I restarted my computer there was space and the program does not met this issue, I guess it might be the virtual memory. |
Beta Was this translation helpful? Give feedback.
-
|
[Promotion of Fork] fenneishi/Fooocus-Control: https://github.com/fenneishi/Fooocus-Control |
Beta Was this translation helpful? Give feedback.
-
|
thanks 🎉 |
Beta Was this translation helpful? Give feedback.
-
2.1.741Style Updates:
Edit: |
Beta Was this translation helpful? Give feedback.
-
|
users who modified |
Beta Was this translation helpful? Give feedback.
-
|
According to my experience in some cases Slightly Cinematic produce much better and realistic Images. You should add it as an option too. |
Beta Was this translation helpful? Give feedback.
-
|
@koltregaskes hi Slightly Cinematic is renamed to "Fooocus Cinematic" |
Beta Was this translation helpful? Give feedback.
-
2.1.750
msedge_xJE5lp57qs.mp4 |
Beta Was this translation helpful? Give feedback.
-
|
right ! .. i used my explorer with directory .. but one important thing 👍
|
Beta Was this translation helpful? Give feedback.
-
|
Great work! Really needed this feature |
Beta Was this translation helpful? Give feedback.
-
|
Bro which Graphics Card do you use? I notice your image generation is very quick. |
Beta Was this translation helpful? Give feedback.
-
|
Hello, friends! Is there any option (maybe something like "Tiling" checkbox or like that) or at least possibility to generate seamless patterns in Fooocus? I've tried several tokens, but they seem to be working not as good as expected. |
Beta Was this translation helpful? Give feedback.
-
2.1.750share a secret - 2.1.750 now can set refiner switch in advanced tab. Just try to use any SD1.5 model as refiner and set refiner switch at 0.4 or 0.5. You will wow. We use special algorithms to achieve this. Do not tell other software that you can do this in fooocus. |
Beta Was this translation helpful? Give feedback.
-
|
the algorithm may change in the future (for example to improve some swapped saturation) so this will not be documented before that |
Beta Was this translation helpful? Give feedback.
-
|
with 0.4 refiner change my emma to someone else (random AI face). |
Beta Was this translation helpful? Give feedback.
-
|
hi all I think I should put an example since it is kind of important I believe Below is "woman, street", seed 1234, run.bat default parameters (default sdxl + refiner)
then you download Realistic Vision V5.1 (it is a SD1.5) https://civitai.com/models/4201/realistic-vision-v51 Then you set it as refiner
Then set refiner switch to 0.4
Then you get this
just omg - what the h
I also tried dreamsharperv8 but Realistic Vision looks better for realistic images |
Beta Was this translation helpful? Give feedback.
-
|
i put a thread here #802 |
Beta Was this translation helpful? Give feedback.
-
|
！！！这就是传说中的1.5与XL混合洗图流 |
Beta Was this translation helpful? Give feedback.
-
|
wow, I'll have to give that a try. Just wondering do SD1.5 Lora's work here? Wondering if they can be mixed and match with SDXL base model. |
Beta Was this translation helpful? Give feedback.
-
|
My graphics card is AMD6700xt and I get an error when I run it,It's:" Device type privateuseone is not supported for torch. Generator() api. " |
Beta Was this translation helpful? Give feedback.
-
|
我的显卡是AMD6700xt,运行时报错Device type privateuseone is not supported for torch.Generator() api.各位大佬有没有啥解决办法，非常感谢！ |
Beta Was this translation helpful? Give feedback.
-
|
Good job! |
Beta Was this translation helpful? Give feedback.
-
|
2.1.754 New Model List |
Beta Was this translation helpful? Give feedback.
-
|
2.1.757 Hi everyone, we are working on some other workloads and we wont update fooocus too much recently, but I spare 20 minutes to impl this
Now you have the image wall. Besides the results are achieved using 2.1.757 default settings. Just "girl in garden" and seed 31337. |
Beta Was this translation helpful? Give feedback.
-
|
I'd like to ask you a question. I made a copy of the language file from the original (English) to translate it into French. But how can I then tell Fooocus to use the translated file? It always defaults to the English file. |
Beta Was this translation helpful? Give feedback.
-
|
see the last sentence of that i18n section on readme |
Beta Was this translation helpful? Give feedback.
-
|
what section i18n? where is it? |
Beta Was this translation helpful? Give feedback.
-
|
https://github.com/lllyasviel/Fooocus#localizationtranslationi18n |
Beta Was this translation helpful? Give feedback.
-
|
I haven't found the file to modify so that it takes into account the fr language file. But I can give you the language file in question for the next update. |
Beta Was this translation helpful? Give feedback.
-
|
Hello. It worked until now, now it throws this error, does anyone know what could be wrong? |
Beta Was this translation helpful? Give feedback.
-
|
Have you tried to reinstall the drive？ |
Beta Was this translation helpful? Give feedback.
-
|
I deleted it from the google drive, I deleted all the processes, I use the shared link, it's still not good |
Beta Was this translation helpful? Give feedback.
-
|
Hi this thread is overflow our maintain capability and please move on to this thread for continued discussion. |
Beta Was this translation helpful? Give feedback.
and others
-
Fooocus has automated the best configuration by default so that users can focus on prompting and generating, and ignore these parameters.
Nevertheless, we also provide a bunch of features for advanced users who are not satisfied by the default. This list will be updated every time we add new features.
Moving from Midjourney to Fooocus
Using Fooocus is as easy as (probably easier than) Midjourney – but this does not mean we lack functionality. Below are the details.
(Unknown method)
(Fooocus has offline GPT-2 based prompt processing engine and lots of sampling improvements so that results are always beautiful, no matter your prompt is as short as “house in garden” or as long as 1000 words)
(Fooocus uses its own inpaint algorithm and inpaint models so that results are more satisfying than all other software that uses standard SDXL inpaint method/model)
(Fooocus uses its own image prompt algorithm so that result quality and prompt understanding are more satisfying than all other software that uses standard SDXL methods like standard IP-Adapters or Revisions)
Fooocus support SDXL models on Civitai
(You can google search “Civitai” if you do not know about it)
Fooocus uses A1111's reweighting algorithm so that results are better than ComfyUI if users directly copy prompts from Civitai. (Because if prompts are written in ComfyUI's reweighting, users are less likely to copy prompt texts as they prefer dragging files)
To use embedding, you can use "(embedding:file_name:1.1)"
We also have a few things borrowed from the best parts of LeonardoAI:
Contents
Sharpness (Added 2023 Aug 15)
Guidance Scale
Fooocus V2 Style (Added 2023 Sep 11)
Sharpness (Added 2023 Aug 15)
Sampling sharpness is developed by Fooocus as a final solution to the problem that SDXL sometimes generates overly smooth images or images with plastic appearance.
Different from other parameters like Automatic1111’s cfg-scale, this sharpness never influences the global structure of images so that it is easy to control and will not mess things up (if you use same random seed). This option is relatively modest so that the image is not likely to be overcooked.
This option can solve 90% problems when the SDXL results are overly smooth.
Example 1
Prompt: a dragon
Default setting. Random seed 12345
Sharpness = 2 (default):
Sharpness = 10:
Sharpness = 20:
Animated Comparison (sharpness 2 vs 10 vs 20):
Example 2
Prompt: a handsome man
Default setting. Random seed 12345
Sharpness = 2 (default):
Sharpness = 10:
Sharpness = 20:
Animated Comparison (sharpness 2 vs 20):
Guidance Scale
The CFG scale from Stable Diffusion community. The difference is that Fooocus implemented an adaptive TSNR rescaling so that results will be less burned when scale is high (like bigger than 15).
Fooocus V2 Style (Added 2023 Sep 11)
Fooocus V2 Style allows users to combine multiple styles at the same time. The combination algorithm of the styles is carefully prepared so that adding multiples styles will not decease the image quality.
Users can often get very high-quality images by combining multiple styles.
Among all those styles the “Fooocus V2” is a very unique dynamic style that use AI (actually it is a GPT2) to adaptively adding styles to make the image more appealing:
The “Fooocus V2” style can be viewed as a smart prompt expansion. We provide examples below.
Note that you may see in console some texts generated by expansion. Those texts are handled in a unique way that is different from your written prompts. In most cases you do not need to worry about these generated keywords influencing your main prompts.
A Quick Comparison for the style "Fooocus V2"
Prompt: Statue of God
Comparison to Midjourney
After the implementation of "Prompt Expansion", the Fooocus can finally compare to Midjourney seriously.
Although their behaviors are quite different still, the results, in many cases, are pretty good and competitive (sometimes even hard to say whose behavior is better).
For example I get a random screenshot of midjourney just now, without any cherry picking, and feed the same prompt to Fooocus default setting:
(This image is really random, I just took a look at that discord without any selections - I trust that Midjourney fans can improve this image A LOT, but here I just pick a random prompt for writing this post)
The prompt I get is: mythical Minotaur, gothic style, ultra realistic, full detail
Fooocus Dark Fantasy:
Apparently Fooocus has different behaviors than Midjourney but we can see that both system is really complicated and the "connection" between prompts and images really went through a "mind journey" (probably a journey inside that prompt expansion model).
Edit Sep 09:
Note that Fooocus Dark Fantasy is here:
Community Forks ...
Fooocus also has many community forks, just like SD-WebUI's vladmandic/automatic and anapnoe/stable-diffusion-webui-ux, for enthusiastic users who want to try!
runew0lf/RuinedFooocus
MoonRide303/Fooocus-MRE
and so on ...
See also About Forking and Promotion of Forks.
Beta Was this translation helpful? Give feedback.
All reactions