Thank you! Hope to get another out in the next couple days. Got distracted today when Magnific AI announced some new features and since I have VIP access I could not resist trying them out. It is great for upscaling generated images.
I’m having fun playing with SD again. Last time was Oct ‘22 so a lot has gotten so much better since. I am interested in the LoRAs, models, and how to use them best along with the various styles. The details you give in your videos is the perfect pace for people to follow along.
honestly sir i'm gratefull people like you exist, you help people out even after a year, i just downloaded fooocus and your tips make it so clear. Also if i may ask you what your specs are? I see you get an image in nearly a minute while i mostly need to wait atleast 2 to 3 for the fastest option and the standard one take me more then 10 minutes, happy holidays. And again thank you for teaching us all this for free.
@@KLEEBZTECH also way i ask which gpu you are using now? Because i'm looking at your text effects video, because i would need something alike what you have been able to do. But it's still lightning fast, and are you still able to do thing on the side?
Introduction and Assumptions - 00:00:01 Launching Fooocus and Preloaded Settings - 00:00:35 File Locations and Folder Structure - 00:01:12 Basic Prompt Usage - 00:02:16 Generating an Image with Basic Prompt - 00:02:44 Adding Details to Prompts - 00:03:28 Using Prompt Weights - 00:04:21 Example of Prompt Weights - 00:05:09 Exploring the Advanced Section - 00:07:04 Performance and Quality Settings - 00:07:04 Aspect Ratios and Number of Images - 00:07:38 Using Negative Prompts - 00:08:12 Understanding and Using Seeds - 00:09:20 Styles Tab Overview - 00:12:32 Applying Styles to Prompts - 00:13:04 Model Tab Overview - 00:14:27 Using Checkpoints and LoRAs - 00:15:07 Example of Using a LoRA - 00:15:47 Guidance Scale and Image Sharpness - 00:18:12 Documentation and Developer Debug Mode - 00:19:51 Conclusion and Future Videos - 00:20:27
I want to ask a question.. so whenever I launch fooocus and generate an image it takes some space on local disk C even though I put fooocus on disk D.. so what does fooocus takes on disk C?
Great video, you earned a sub! Could you make a video how you can train a Lora? Let's say I want to create certain scenes with specific characters. I was thinking to do it with face swap but this might be a better option. Thanks
I would like to at some point. I currently do not have enough VRAM to do it locally but am looking at options. But I will say that FaceSwap can be decent. I have been experimenting with it to figure out how to get best results. I probably will share something very soon.
I have not. My normal use for these tools does not require me to deal with faces matching so I have not played around with a ton of it. Have been trying things lately. @@billseigs
I have nvidia gtx 1050 4 gb ram with 8gb ram and foocus doesn’t work .cannot generate photos .roght after installation when i enter prompt in the Ui ,it gives me error . Any advice ??
I don't think your system will be able to run it locally. But you can go through this and see if anything helps github.com/lllyasviel/Fooocus/blob/main/troubleshoot.md
In the image promt (and I've checked face swap). What does "weight" do? "stop at" is to respect more the original image prompt (higher value) I think I know.
Think of weight like a volume knob. The higher it is the more it influences the image. Stop at is when it stops influencing. So a .5 on stop at means that it will no longer use that image during the generation prompt at 50% of the way.
Fast upscale should but it won't really add much for details. But if you combine Faceswap with the other two upscale options it should also do a decent job of keeping the face and adding more details. I have had decent luck. But you do need to enable the option in debug options to use faceswap with the upscales.
a nice turtorial with clear instructions and explains, awesome job for that. i love the sharpness to set on 7-10 if i make cartoon/anime/Stylized stuff because the background, muscles, clothes, eyes(eyelashes etc.) or the skin have much more details, really amazing. but i have a question about the prompts? i have sometimes this feeling that Fooocus can handle better with short words as with a detail explanation. if i write "a woman dance with a man, dancing, ball room, big window........" it is okay but if i write more details like " a woman with green eyes and dark toned skin dance with a Tall man....more details like Clothes, hair, blablabla" fooocus dont know how to handle this prompts. gives a trick how i can write detail characters with different features and backgrounds? (sorry for my bad english)
Usually the more complex a prompt the more likely things will get mixed up. You want to keep it as simple as possible for best results usually. I do plan on making a video on prompts in the near future but I myself have not found the perfect solution to that. Often I use inpainting to fix certain parts of images if it get minor things wrong. Hopefully that will get better in the future. Some of the other ones like Midjourney and DALL-E have gotten much better at understanding prompts so I am hoping for the same. I do have lots of tips but trying to put them in comments would be hard. I do suggest things like using weights and also remember that things at the beginning of the prompt will have more weight as well.
Also you could consider using the image prompt feature. You can use that to influence the text prompt and reduce the need for too many details. I am not sure if you are familiar with that and if not I do have a video on that. ruclips.net/video/2hW0M-Y1R6k/видео.html
@@KLEEBZTECH i watched your videos and this video is the most of it now because you are the only one that can explain what the differnce between the stop at and weight section.
I do a lot of fantasy "realism" and one of the hardest things I keep running into, is "human ears" keep forever showing up on things that have cat or dog or bunny .. you get the idea, is there a surefire way to generate a picture with "no human ears" ?
There are multiple ways. One put 'full body' in prompt. Also could mention type of footwear or bare feet. Another thing is to use an aspect ratio that is taller.
The original SDXL needed the use of a refiner for better quality. For most checkpoints you do not need to use a refiner but you can if you want. Some people will use it to combine the different capabilities of each checkpiont. So maybe you start with anime and then for the last 20% you switch to a photographic geared one to get a different look.
Not normally. I have not tested myself but from what others have said you can use an SD 1.5 checkpoint as a refiner which would allow you to use 1.5 Loras. Also would want it to switch to the refiner early if doing that. My understanding that is the only work around.
i am get this error"TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()" any idea how to solve'
Truly the 'GOAT' of Fooocus!
Thank you!
Tego the GOAT is the real GOAT. 😀
I’m on the edge of my seat. Part three please. Keep up the good work you’re killing it.
Thank you! Hope to get another out in the next couple days. Got distracted today when Magnific AI announced some new features and since I have VIP access I could not resist trying them out. It is great for upscaling generated images.
I’m having fun playing with SD again. Last time was Oct ‘22 so a lot has gotten so much better since. I am interested in the LoRAs, models, and how to use them best along with the various styles. The details you give in your videos is the perfect pace for people to follow along.
bro your teaching way is amazing
Thanks a ton.
agree
btw, if you want to change the style of the same image, use the image prompt and no text prompt. This way it keeps the basic image composition.
Thanks for clarifying about the prompts weights, that was very educational! As well as the seed and how to check it out!
honestly sir i'm gratefull people like you exist, you help people out even after a year, i just downloaded fooocus and your tips make it so clear. Also if i may ask you what your specs are? I see you get an image in nearly a minute while i mostly need to wait atleast 2 to 3 for the fastest option and the standard one take me more then 10 minutes, happy holidays. And again thank you for teaching us all this for free.
I was using a 3070 with 8GB VRAM at the time. Your GPU is the major factor with these.
@@KLEEBZTECH thank you so much for your quick response, i will change my gpu then.
@thomasvdm2527 the more VRAM the better.
@@KLEEBZTECH also way i ask which gpu you are using now? Because i'm looking at your text effects video, because i would need something alike what you have been able to do. But it's still lightning fast, and are you still able to do thing on the side?
@thomasvdm2527 I am using a 4070 with 12GB VRAM now. I do wish I had even more VRAM. I do often speed up the parts in the videos when generating.
Everything i learned about Fooocus its from your teachings.
Thank you very much Sir
starting this amazing serie! great explanations with examples, i love ur content!
Thank you!
thanks mate for teaching these valuable lessons for free
My pleasure!
Introduction and Assumptions - 00:00:01
Launching Fooocus and Preloaded Settings - 00:00:35
File Locations and Folder Structure - 00:01:12
Basic Prompt Usage - 00:02:16
Generating an Image with Basic Prompt - 00:02:44
Adding Details to Prompts - 00:03:28
Using Prompt Weights - 00:04:21
Example of Prompt Weights - 00:05:09
Exploring the Advanced Section - 00:07:04
Performance and Quality Settings - 00:07:04
Aspect Ratios and Number of Images - 00:07:38
Using Negative Prompts - 00:08:12
Understanding and Using Seeds - 00:09:20
Styles Tab Overview - 00:12:32
Applying Styles to Prompts - 00:13:04
Model Tab Overview - 00:14:27
Using Checkpoints and LoRAs - 00:15:07
Example of Using a LoRA - 00:15:47
Guidance Scale and Image Sharpness - 00:18:12
Documentation and Developer Debug Mode - 00:19:51
Conclusion and Future Videos - 00:20:27
Thanks Dude for uploading this tutorial.👍💜
Welcome 😊
"...big anything" lol, understood!
I want to ask a question.. so whenever I launch fooocus and generate an image it takes some space on local disk C even though I put fooocus on disk D.. so what does fooocus takes on disk C?
What is a difference between lora and checkpoint?! Whether is the checkpoint can be entered?!
I click on run realist and it shows press any key and then it disappears happens with anime and bat as well please help me get through it
which configuration graphics card u used? and what is ur computer whole configuration??
I have an i5 with 32GB RAM and 3070 with 8GB VRAM.
Great video, you earned a sub! Could you make a video how you can train a Lora? Let's say I want to create certain scenes with specific characters. I was thinking to do it with face swap but this might be a better option.
Thanks
I would like to at some point. I currently do not have enough VRAM to do it locally but am looking at options. But I will say that FaceSwap can be decent. I have been experimenting with it to figure out how to get best results. I probably will share something very soon.
@@KLEEBZTECHThank you, it will be much appreciated! Have you tried Reactor for face swapping? What do you think of it?
I have not. My normal use for these tools does not require me to deal with faces matching so I have not played around with a ton of it. Have been trying things lately. @@billseigs
I have nvidia gtx 1050 4 gb ram with 8gb ram and foocus doesn’t work .cannot generate photos .roght after installation when i enter prompt in the Ui ,it gives me error .
Any advice ??
I don't think your system will be able to run it locally. But you can go through this and see if anything helps github.com/lllyasviel/Fooocus/blob/main/troubleshoot.md
Why can't I increase the weight with ctrl +up? Is there a setting I have to activate first?
It should work by default, there is no setting for it. It is possible that it could be a browser issue or something maybe.
why there is no 16:9 aspect ration?
you are the best man, i mena it
Thanks! Check out some of the other videos I have on Fooocus for getting even more out of it.
In the image promt (and I've checked face swap). What does "weight" do? "stop at" is to respect more the original image prompt (higher value) I think I know.
Think of weight like a volume knob. The higher it is the more it influences the image. Stop at is when it stops influencing. So a .5 on stop at means that it will no longer use that image during the generation prompt at 50% of the way.
Thanks! It's posible to upscale an image and keep the same exactly? (I choose upscale 2x but it changes de characters)@@KLEEBZTECH
Fast upscale should but it won't really add much for details. But if you combine Faceswap with the other two upscale options it should also do a decent job of keeping the face and adding more details. I have had decent luck. But you do need to enable the option in debug options to use faceswap with the upscales.
I do cover faceswap in more detail in my other videos as well.
@@KLEEBZTECH I tried that, but it throw me really bad results (puts a woman face on mine for example). I'll watch your video and try again...
Thank you ❤
a nice turtorial with clear instructions and explains, awesome job for that. i love the sharpness to set on 7-10 if i make cartoon/anime/Stylized stuff because the background, muscles, clothes, eyes(eyelashes etc.) or the skin have much more details, really amazing.
but i have a question about the prompts?
i have sometimes this feeling that Fooocus can handle better with short words as with a detail explanation.
if i write "a woman dance with a man, dancing, ball room, big window........" it is okay but if i write more details like " a woman with green eyes and dark toned skin dance with a Tall man....more details like Clothes, hair, blablabla" fooocus dont know how to handle this prompts. gives a trick how i can write detail characters with different features and backgrounds?
(sorry for my bad english)
Usually the more complex a prompt the more likely things will get mixed up. You want to keep it as simple as possible for best results usually. I do plan on making a video on prompts in the near future but I myself have not found the perfect solution to that. Often I use inpainting to fix certain parts of images if it get minor things wrong. Hopefully that will get better in the future. Some of the other ones like Midjourney and DALL-E have gotten much better at understanding prompts so I am hoping for the same. I do have lots of tips but trying to put them in comments would be hard. I do suggest things like using weights and also remember that things at the beginning of the prompt will have more weight as well.
@@KLEEBZTECH the inpaint mode is really helpful, thank you for the help and your videos are really good.
Also you could consider using the image prompt feature. You can use that to influence the text prompt and reduce the need for too many details. I am not sure if you are familiar with that and if not I do have a video on that. ruclips.net/video/2hW0M-Y1R6k/видео.html
@@KLEEBZTECH i watched your videos and this video is the most of it now because you are the only one that can explain what the differnce between the stop at and weight section.
I do a lot of fantasy "realism" and one of the hardest things I keep running into, is "human ears" keep forever showing up on things that have cat or dog or bunny .. you get the idea, is there a surefire way to generate a picture with "no human ears" ?
You could try putting ears in the negative with added weight.
awesome! what's your gpu?
I was using an Nvidia 3070 but not using a 4070 with 12 GB VRAM. Much better.
Hi, how to generate full body image using Fooocus? Please guide
There are multiple ways. One put 'full body' in prompt. Also could mention type of footwear or bare feet. Another thing is to use an aspect ratio that is taller.
Hello please send me the collab link where i can run it on coolab directly
It is on the Github page. github.com/lllyasviel/Fooocus?tab=readme-ov-file#colab
Hey buddy, what is the use of base model and refiner?!! You have rushed that concept in this video..anyone please explain
The original SDXL needed the use of a refiner for better quality. For most checkpoints you do not need to use a refiner but you can if you want. Some people will use it to combine the different capabilities of each checkpiont. So maybe you start with anime and then for the last 20% you switch to a photographic geared one to get a different look.
@@KLEEBZTECH thanks !!
Can you use SD1.5 Loras?
Not normally. I have not tested myself but from what others have said you can use an SD 1.5 checkpoint as a refiner which would allow you to use 1.5 Loras. Also would want it to switch to the refiner early if doing that. My understanding that is the only work around.
brother i am having file issue
What issues? Need to be a little more specific.
GOAT
👍👍👍👍
i am get this error"TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()"
any idea how to solve'
That is just a warning and should not be causing any issues with Fooocus. github.com/lllyasviel/Fooocus/discussions/3465
w dól leci