If you find the video useful and would like to tip, RUclips does have the Super Thanks option. Those are greatly appreciated!!! ☕ Or you can buy me a coffee or see what I might be working on: ko-fi.com/kleebztech
From this video, I learned new things, such as the fact that watermarked and blurred photos are not ideal for image prompts, especially when imitating poses. As mentioned in the final part of the video, we should delve deeper into the more advanced aspects of Focus, since it's already easy enough to use. I don't mind if the video is two hours long; I will definitely watch it. In the next one, please cover Loras/checkpoints in Focus. Thank you for the video; it's been really helpful
Thank you! I was having a ton of trouble figuring out how the various advanced options worked especially when combined. The documentation out there is very lacking.
Thank you for your hard work, this is a great video! I noticed that the checkpoint also plays a role - please share the information - which model did you use for the examples in this video? Thanks!
Yes, I was wondering what the PyraCanny & the CPDS was all about. I will give them a go. I have had much experience with the face-swap function and very much like that. I didn't consider the idea of using two face-swap example images of the same model. That's an interesting idea. What I have done though is experimented with two different models for images with two characters in them. That becomes a little tricky. The balance of the Stop & Weight becomes critical, and the text prompts also become "delicate" and easily broken. And despite that it will sometimes morph both examples into one composite character. Which can be interesting/hilarious if one is male & the other female. Another couple of things I have found is that the face-swap example doesn't necessarily need to be head & shoulders only. And a full body or total upper body photo can help achieve a consistency in body type too. Having a hi res image is definitely important. Also, when you have an Asian model, it has a tendency to "Europeanise" them. You can up the weight etc, but what seems to be more effective is simply including the word "Asian" in the text prompt.
One downfall of using an image that is not just the headshot is the other stuff in the image will influence things I find. Especially if only one image. For example if wearing a certain item that will keep showing up. One advantage I have found of using multiple head shots with varying backgrounds is less influence from just one. But still experimenting with other ideas.
@@KLEEBZTECH Yes, definitely it will. Which can be an advantage or disadvantage depending on what you are trying to create. I tried using the multiple angles of the same model that you suggested and it worked really well. I reckon I'll keep doing that moving forward. And I used the Pyrocanny to finally get my model to wear a spiky bra! A picture is worth 1000 words as they say. Tried all sorts of ideas with text prompts, even got GPT to write some but there just weren't enough spikes on the bra! Bringing in an image with Pyrocanny fixed that.
Very comprehensive, thanks a lot! Several questions - what does the Weight plugin do, what is "Weight" in Fooocus per se? Realism? Graphical quality? Depth / perspetive? Also, is there a way to just generate a new face on an already uploaded photo, and if yes is it done the same way as in the faceswap feature?
The best way to think of weight is like volume. The higher it is the more influence it has. If you want to swap faces on an already generated photo then you can do the inpainting with face swap. Do need to turn on the option in debug tools to enable Faceswap and inpainting together. I do cover it more in some of my other videos like this one: ruclips.net/video/MntZa4qLwn8/видео.htmlsi=voH7TodZPFPJ5m-h&t=603
Dunno if you're still active on this channel but I've been learning a lot from your videos. I'd like to ask you that whenever I am using pyracanny with image prompt the generated images have very soft lighting even though both of the images have bright lights. Your help would be deeply appreciated
Pyracanny really should not have much of an impact on that from my experience since it is just using the structure of the image to guide things. Are you using multiple images in the image prompt? Are both set to pyracanny? How about if just one?
@@KLEEBZTECH hey so nice of you to assist in this. I have one image in pyracanny and the other one in image prompt. How can I improve the lights without losing cloth colors
Without seeing what you are working with it is hard to say since what you have described I don't recall running into. If a lighting issue and you think it has to do with the image prompts I would test with maybe just the pyracanny only first and then with just the image prompt one. Of course the other thing is to add more lighting details to the text prompt. Also may want to try adjusting the settings to see if any difference.
I have one question please. BTW, thank you so much for your great, simple English explanations about FOOOCUS. I want to do an exact face swap from a face image, and paste that exact face (with as little variation as possible) onto another full body image (obviously with its own face). I haven't been able to get anything close to duplicating the face. How would you suggest I do that?
I do have a couple other videos that cover more on Faceswap. The first big thing is you need a good quality reference image of the face and in the image prompt section check the advanced at the bottom and use Faceswap with about .95 stop and and .9 weight. There are several ways of doing it. Inpainting is most likely how you will need to do it. Need to enable in debug inpainting and image prompt together. And some images it is harder than others.
@@KLEEBZTECH Thanks so much for taking the time to answer. I'm going to check for your other videos right now. BTW I subbed a few months ago. Love your videos.
@@iheartjetaviation2764 Thanks for the sub! And I will say I do not have all the answers when it comes to Faceswap and a lot just comes down to tweaking the settings and the source images. Good luck.
Is there any way to use a character as a prompt and merge into a background? Let's say - as an example - I have a photo of a field with flowers and another picture of a guy that walks. Can I merge the two? Can I make the guy walk in the field? Thank you.
at 21:15 You have mentioned that it is possible to use multiple images for face swapping. Would that improve accuracy? Perhaps you could dedicate a video solely to that, as well as to other different techniques or options for face swapping in Fooocus.
I have not tested it fully yet myself but from what I have researched it should improve the accuracy. I do plan on a more extensive video on that once I get more of the basic stuff out of the way and get enough test generations done.
Those are really fast generation! Am on 3060 12gb and it takes quite a while when using quality. Was your video sped out during generation? Would you mind sharing your build or GPU for this? Thanks!
Mine probably takes 30 to 45 seconds per generation I think but I don't really pay that close attention. I'm usually recording and everything else and it taxes my system. Certain things I need to actually stop recording when it needs to upscale especially.
@@KLEEBZTECHI use in Google colab, but not able to process multiple images, ( like heavy task ) GPU disconnecting again and again on colab! 😭, I need this for my AI Influencer!!!!
If you just wanted to change everything else in the image and leave the character alone you could use inpainting possibly for what you are looking for but not sure if that is what you mean. If you mean keeping the same body characteristics then that might be something to do with a trained lora on one person.
@@KLEEBZTECH will you make a video about how to train a lora? it would be cool and usefull. i wouldl ove it :>. as always, fantastic explaination and examples. im eating this serie following you step by step with my images. I tried to use FaceSwap (for face) and CPDS (for pose) but it changes the face, a lot. both images are good quality, i also tried to upscale2x them, but it didnt change :
Any suggestions on how to generate images that form a sequence of actions, like a character standing up or grabbing something. Or how to accurately move or rotate specific limbs?
This might give you some ideas. I have no perfect solution but if you can get it to create the poses into a grid of 4 images then you can often get the same character in each image. Moving or rotating limbs you could try doing with inpainting but that may be difficult to get it to look right depending on the image. ruclips.net/video/xGkGPbNjjr0/видео.html
I thought there was a FOOOCUS for photos that doesn't use stable diffusion. There's plenty face swap for photos, but they are all online. Why has it become so difficult to get anything offline anymore? Does anyone know of a total offline app?
@@KLEEBZTECH I hope to use it for making images for the advertising agency I work for. Having consistent characters would be helpful to realize this. Have you found a good way to achieve this in Fooocus?
I have the question about GPU - does the weaker GPU with lower amount of RAM only affects the time of the image processing or it also affects the quality of the output image?
Well this does cover the advanced settings but did not dive into some of the hidden debug tools. Yes you should be able to use most things with the image prompt feature like loras. But the results may not always be the expected results depending on the lora.
Really depends on what you are doing. I find Fooocus is best for most of what I do. Some things I have to use other tools and I also like to keep familiar with them, but really like the ease of use and the inpainting works really well.
I will have to look into it more since by default SD does not do well with text. I usually have edited after with traditional tools when I need text. I am hoping that it will get better since other tools like Midjourney have made some major improvements.
For me bat file runs just fine and it opens in a web browser but my computer quickly starts to heat up:(. Maybe if I turned my resolution down as much as possible things might work better:(
Is there any way to do this; youve got an image with multiple characters but you want change the face of one person in that image and keep the face of the rest of the chatacters in the image
This has the information for Macs. Seems they are not fully tested but can work. I do not have a Mac to test. github.com/lllyasviel/Fooocus?tab=readme-ov-file#mac
If you want to use multiple reference images for face swap you turn them all on face swap that you want. I also find that cranking up the stop at and weight to almost one each is usually where I find my best luck. But also need good reference pictures
I was doing some testing today and using some personal pictures of family because I will definitely notice if it looks correct or not. It was still pretty hit or miss and I think it mostly depends on reference pictures. Sometimes one can be better.
Genius asks. In Fooocus, put together an image of a woman with certain characteristics. Of all the ones he generated, one is the one I liked the most. I use Image Prompt and Face Swap with that photo so that they all come out with the same face, let's say so that it is the same person. It FUNCION. All the photos I order that way are the same person. But I see that by using face swap to be the same person doing different things, it limits a lot. Sometimes he uses the same tones from the original photo that he used in Face Swap, meaning if I ask for Bikini he makes it blue like Jean's jacket in the original photo and many things I ask for are limited by this original photo. If I don't use Image Prompt and ask to generate what you want you have much more freedom in creating. So the question is, can I generate without face swap to have all the freedom and then apply the face swap to that photo that I generated so that it is always the same person I want? That is, for example, put a 23-year-old blonde woman with green eyes, drinking coffee on that side, the photo is generated for me, but the face that was generated for me is nothing like the one I always have to use with Face Swap and that generated photo. give him the face I want? I don't know if I explain myself. I'll give you my Instagram so you can see the character that is always made with the same photo as I said with Image Prompt and Face Swap but this way I always get the same face or the same person but the results it gives me are greatly limited. @emilia_diaz_fit So i need to know the best way to generare the same person doing different things in differents scenarios without being limited by the photo that i used to face swap. Is for a ig character it need to be the same person. Thank you so much
Often I speed up generation in my videos but normally mine take about 30 seconds to 40 seconds depending on what I'm doing. I'm using a 3070 with 8 GB of vram. That is the biggest factor.
@@KLEEBZTECH pfff, that's nasty.. Anyway, can you make a video or something about how to actually write a prompt to be followed? For example I'm creating a model and I keep writing that I want her with red eyes high contrast and nose piercing and I never get those. Maybe there are some tips and tricks idk to write as keywords for better results
I do plan on a video about prompting but will admit I do not consider myself great at that either. Still much of it is luck since prompt adherence is not great with these. They do say SD 3 will be much better. First thing I will mention is to disable styles to start with. They can sometimes cause issues. Masterpiece for example has the word earrings included. Also make use of inpainting if it doesn't get one or two things. Especially since it takes you so long to generate it can save an almost perfect image. Colors can be difficult since they can impact the wrong thing in the image. With inpainting you can tweak the prompt as you modify. Also be sure to use extra weight for the words or phrases you want to stress if needed. (nose piecing:1.3) for example. Hope those tips help if you were not already doing do them.
Are you running locally or through something like colab? If the first time using it will download what is needed and if you have connection issues that might also be what is wrong. I am not sure.
The problem is that FaceSwap is going to try to make those faces turn towards the camera. You might be able to get it to do it a little but that is a challenge. I have not really experimented much with trying to get it to honestly.
do you have colab pro? because it tells me i dont have nvidia gpu. am i run out of free gpu credit? cause it gives error instead of producing the images.
You are not saying how you are using it. I am going to assume you are using it online somewhere? Are you using on Colab? If so then Colab only lets you use so much of their computer power each day for free. Then you have to wait until the next day or pay for the paid plan with Colab.
stable fusion -> forge. all same and less simpler gui. we allready have all setup xD. only need correct model and prompt todo your image. yes if model is strained what you want. it create 0% its all trained
If you find the video useful and would like to tip, RUclips does have the Super Thanks option. Those are greatly appreciated!!! ☕ Or you can buy me a coffee or see what I might be working on: ko-fi.com/kleebztech
Amazing, I was confused with the CPDS, but it now clarified, thank you very much
You're very welcome! I'm currently working on an updated video for poses. Hopefully it will be uploaded in the next day or two.
Hey i just wanted to say thank you. you have a real calming voice and explain every detail of the whys and why nots really well. Thank you for that
Very welcome.
Finally a tutorial on how to use Fooocus in depth, thank you, I subscribe and like, I hope for more videos related to Fooocus, thank you very much
Many more are planned. Just posted one on inpainting. 👍
Finally a tutorial i was looking for.tysm
this was very useful, thank you so much Kleebz!
You're very welcome!
This is fantastic. Thank you for your videos!
Glad you enjoyed it!
Extremely thankful for guiding me through all these options, much love from India (subbed)
Awesome! Thank you!
Man, this eats memory hard. I have a decent GPU (6800 XT with 16GB VRAM) but 16GB RAM makes each image take like 2 minutes. I still love it though
Best Tutorial Error For Fooocos! You Earned A Subscriber Bro!
Thanks! Don't miss some of my other videos. Lots of tricks to using Fooocus in them.
"Thank you, you explained it very deeply and in an easily understandable way. Your teaching skill is very good, without wasting a single word."
You're very welcome! Thank you for the feedback.
something very sus about such a positive comment under an AI video still with quotation marks around it...
@@opiwaran354 😂
Thanks! Peace & Blessings 😁
Thank you as well! Really appreciate it.
From this video, I learned new things, such as the fact that watermarked and blurred photos are not ideal for image prompts, especially when imitating poses.
As mentioned in the final part of the video, we should delve deeper into the more advanced aspects of Focus, since it's already easy enough to use. I don't mind if the video is two hours long; I will definitely watch it.
In the next one, please cover Loras/checkpoints in Focus. Thank you for the video; it's been really helpful
Thank you! I was having a ton of trouble figuring out how the various advanced options worked especially when combined. The documentation out there is very lacking.
Glad it helped! There are definitely many ways of mixing and matching things. I'm still learning. And some of it is still luck.
Thanks for the explanation guide. I was having a time figuring out how to use those tools. This was of great help for me. 👍
You're welcome! I'm still trying to learn more advanced stuff and plan on posting more videos as I discover more
Thanks! I learned a few things here! Props to YOU
Glad it was helpful!
Thanks for sharing 😊
Thanks for watching!
Great. Yes, PyraCanny it's strokes, but CPDS works with depth, not just decolorize.
Thank you for your hard work, this is a great video! I noticed that the checkpoint also plays a role - please share the information - which model did you use for the examples in this video? Thanks!
I used the default which is what I usually use unless I mention otherwise. That would have been JuggernautV8.
Great video,, Thank you !
Thank you for the good work bro
No problem 👍
Yes, I was wondering what the PyraCanny & the CPDS was all about. I will give them a go. I have had much experience with the face-swap function and very much like that. I didn't consider the idea of using two face-swap example images of the same model. That's an interesting idea.
What I have done though is experimented with two different models for images with two characters in them. That becomes a little tricky. The balance of the Stop & Weight becomes critical, and the text prompts also become "delicate" and easily broken. And despite that it will sometimes morph both examples into one composite character. Which can be interesting/hilarious if one is male & the other female.
Another couple of things I have found is that the face-swap example doesn't necessarily need to be head & shoulders only. And a full body or total upper body photo can help achieve a consistency in body type too. Having a hi res image is definitely important. Also, when you have an Asian model, it has a tendency to "Europeanise" them. You can up the weight etc, but what seems to be more effective is simply including the word "Asian" in the text prompt.
Good tips!
One downfall of using an image that is not just the headshot is the other stuff in the image will influence things I find. Especially if only one image. For example if wearing a certain item that will keep showing up. One advantage I have found of using multiple head shots with varying backgrounds is less influence from just one. But still experimenting with other ideas.
@@KLEEBZTECH Yes, definitely it will. Which can be an advantage or disadvantage depending on what you are trying to create. I tried using the multiple angles of the same model that you suggested and it worked really well. I reckon I'll keep doing that moving forward. And I used the Pyrocanny to finally get my model to wear a spiky bra! A picture is worth 1000 words as they say. Tried all sorts of ideas with text prompts, even got GPT to write some but there just weren't enough spikes on the bra! Bringing in an image with Pyrocanny fixed that.
This video is fantastic!
Thank you very much.
Awesome! New sub =). Please keep demonstrating Fooocus in depth. Maybe tweaking on 'Developer Debug Mode'. Thank you!
Thanks! Will do.
Thanks best tutorial i've seen so far. Faceswap still not as good as ReActor on 1.5 but im sure in the future it'll improve.
Yeah Faceswap I don't really use often and hope for some improvements.
Very comprehensive, thanks a lot! Several questions - what does the Weight plugin do, what is "Weight" in Fooocus per se? Realism? Graphical quality? Depth / perspetive? Also, is there a way to just generate a new face on an already uploaded photo, and if yes is it done the same way as in the faceswap feature?
The best way to think of weight is like volume. The higher it is the more influence it has. If you want to swap faces on an already generated photo then you can do the inpainting with face swap. Do need to turn on the option in debug tools to enable Faceswap and inpainting together. I do cover it more in some of my other videos like this one: ruclips.net/video/MntZa4qLwn8/видео.htmlsi=voH7TodZPFPJ5m-h&t=603
@@KLEEBZTECH Appreciate it! Great tutorial!
Very, very helpful, thank you.
You're welcome!
Dunno if you're still active on this channel but I've been learning a lot from your videos.
I'd like to ask you that whenever I am using pyracanny with image prompt the generated images have very soft lighting even though both of the images have bright lights.
Your help would be deeply appreciated
Pyracanny really should not have much of an impact on that from my experience since it is just using the structure of the image to guide things. Are you using multiple images in the image prompt? Are both set to pyracanny? How about if just one?
@@KLEEBZTECH hey so nice of you to assist in this. I have one image in pyracanny and the other one in image prompt. How can I improve the lights without losing cloth colors
Without seeing what you are working with it is hard to say since what you have described I don't recall running into. If a lighting issue and you think it has to do with the image prompts I would test with maybe just the pyracanny only first and then with just the image prompt one. Of course the other thing is to add more lighting details to the text prompt. Also may want to try adjusting the settings to see if any difference.
@@KLEEBZTECH I'll try with adding lighting prompt, I am trying to have one pictures outfit to the other
Do you have Instagram where I can reach you?
@@KLEEBZTECH sending you a request right away. I am new to discord though
Thanks for the video! One question, what video card are you using? I use a 6GB GPU but it is very slow.
3070 with 8GB. 60 steps quality setting it takes about 42 seconds.
I have one question please. BTW, thank you so much for your great, simple English explanations about FOOOCUS. I want to do an exact face swap from a face image, and paste that exact face (with as little variation as possible) onto another full body image (obviously with its own face). I haven't been able to get anything close to duplicating the face. How would you suggest I do that?
I do have a couple other videos that cover more on Faceswap. The first big thing is you need a good quality reference image of the face and in the image prompt section check the advanced at the bottom and use Faceswap with about .95 stop and and .9 weight. There are several ways of doing it. Inpainting is most likely how you will need to do it. Need to enable in debug inpainting and image prompt together. And some images it is harder than others.
@@KLEEBZTECH Thanks so much for taking the time to answer. I'm going to check for your other videos right now. BTW I subbed a few months ago. Love your videos.
@@iheartjetaviation2764 Thanks for the sub! And I will say I do not have all the answers when it comes to Faceswap and a lot just comes down to tweaking the settings and the source images. Good luck.
Is there any way to use a character as a prompt and merge into a background? Let's say - as an example - I have a photo of a field with flowers and another picture of a guy that walks. Can I merge the two? Can I make the guy walk in the field? Thank you.
super helpful! thanks alot!
You're welcome!
thanks
Welcome
How can i swap the face and let him get the pose of an other image ?
You would set one image prompt to FaceSwap and another to either Pyracanny or CPDS for the pose.
at 21:15 You have mentioned that it is possible to use multiple images for face swapping. Would that improve accuracy? Perhaps you could dedicate a video solely to that, as well as to other different techniques or options for face swapping in Fooocus.
I have not tested it fully yet myself but from what I have researched it should improve the accuracy. I do plan on a more extensive video on that once I get more of the basic stuff out of the way and get enough test generations done.
If I can convince someone to trade a 4090 for my 3070 it will happen sooner. lol.
Those are really fast generation! Am on 3060 12gb and it takes quite a while when using quality. Was your video sped out during generation? Would you mind sharing your build or GPU for this? Thanks!
Oh absolutely sped up. Time to go grab the coffee. 😁 I am running an i5 a couple generations back. 32gb RAM. Nvidia 3070 with 8gb VRAM.
Mine probably takes 30 to 45 seconds per generation I think but I don't really pay that close attention. I'm usually recording and everything else and it taxes my system. Certain things I need to actually stop recording when it needs to upscale especially.
@@KLEEBZTECHI use in Google colab, but not able to process multiple images, ( like heavy task ) GPU disconnecting again and again on colab! 😭, I need this for my AI Influencer!!!!
I don’t want to swap just the face. I want to swap the whole character, both face & body included. I wonder if that’s possible with any AI tool.
If you just wanted to change everything else in the image and leave the character alone you could use inpainting possibly for what you are looking for but not sure if that is what you mean. If you mean keeping the same body characteristics then that might be something to do with a trained lora on one person.
@@KLEEBZTECH will you make a video about how to train a lora? it would be cool and usefull. i wouldl ove it :>. as always, fantastic explaination and examples. im eating this serie following you step by step with my images. I tried to use FaceSwap (for face) and CPDS (for pose) but it changes the face, a lot. both images are good quality, i also tried to upscale2x them, but it didnt change :
Any suggestions on how to generate images that form a sequence of actions, like a character standing up or grabbing something. Or how to accurately move or rotate specific limbs?
This might give you some ideas. I have no perfect solution but if you can get it to create the poses into a grid of 4 images then you can often get the same character in each image. Moving or rotating limbs you could try doing with inpainting but that may be difficult to get it to look right depending on the image. ruclips.net/video/xGkGPbNjjr0/видео.html
Cool! Thanks!
Welcome!
Thanks!
Very welcome!
I thought there was a FOOOCUS for photos that doesn't use stable diffusion. There's plenty face swap for photos, but they are all online. Why has it become so difficult to get anything offline anymore? Does anyone know of a total offline app?
Nice video. Am still searching for a use-case for this other then wowing myself at what's possible. 😂
I found once I understood how it worked I use the image prompt feature much more than I did in the past. I'm always finding new ways of utilizing it.
@@KLEEBZTECH I hope to use it for making images for the advertising agency I work for. Having consistent characters would be helpful to realize this. Have you found a good way to achieve this in Fooocus?
My last couple videos go over some tips for consistent characters. ruclips.net/video/Dt9Jfocv4dc/видео.html
@@KLEEBZTECH thanks for the reply. 🙂 ill check these out tomorrow
Just added another one as well. Hopefully they will give you some ideas.
I have the question about GPU - does the weaker GPU with lower amount of RAM only affects the time of the image processing or it also affects the quality of the output image?
That should have no impact on the quality of the image.
Would love to see the advanced settings! Can you use loras as well?
Well this does cover the advanced settings but did not dive into some of the hidden debug tools. Yes you should be able to use most things with the image prompt feature like loras. But the results may not always be the expected results depending on the lora.
ComfyUI still the best
Really depends on what you are doing. I find Fooocus is best for most of what I do. Some things I have to use other tools and I also like to keep familiar with them, but really like the ease of use and the inpainting works really well.
But ComfyUI I prefer well over A1111 when it comes to doing other stuff.
thank you so useful, could you make a video about adding titles and texts?
I will have to look into it more since by default SD does not do well with text. I usually have edited after with traditional tools when I need text. I am hoping that it will get better since other tools like Midjourney have made some major improvements.
For me bat file runs just fine and it opens in a web browser but my computer quickly starts to heat up:(. Maybe if I turned my resolution down as much as possible things might work better:(
If your computer is actually heating up I would make sure you blow out any dust. Also make sure your fans work
Great vid. I swear there must be a way to apply style A to picture B without changing the content of picture B but I'll be damned if I know it!
I downloaded this version, but inhave no idea about this. Is better automatic1111? Or it is a totally diferent thing
I would not say better but it is much easier to use.
@@KLEEBZTECH how can i change the model? Normally i have the same kind of face, i normally change it using prompt imagine with other face.
👍👍👍👍👍
Is there any way to do this; youve got an image with multiple characters but you want change the face of one person in that image and keep the face of the rest of the chatacters in the image
You would need to use inpainting. I do have some videos on that. Also need to enable the feature of mixing inpainting and faceswap.
@@KLEEBZTECH will try that and get back to you with the results. Thanks man
for better poses shift from speed to quality
Good to know. I almost always use the quality setting so have not noticed.
Also is it possible to get exactly same clothing in every generation?
No.
is there a way to keep the same style and only change the pose?
Pose is only controlled by pyracanny or CPDS so you just need to use those for pose.
Do img prompt amd canny come built in with fooocus?
Yes the things that I showed are all part of Fooocus.
When will it be available for the iMac?
This has the information for Macs. Seems they are not fully tested but can work. I do not have a Mac to test. github.com/lllyasviel/Fooocus?tab=readme-ov-file#mac
on the multiple face swaps. do you put on face swap or pyracanny?
If you want to use multiple reference images for face swap you turn them all on face swap that you want. I also find that cranking up the stop at and weight to almost one each is usually where I find my best luck. But also need good reference pictures
I was doing some testing today and using some personal pictures of family because I will definitely notice if it looks correct or not. It was still pretty hit or miss and I think it mostly depends on reference pictures. Sometimes one can be better.
thanks! yes i have been playing with it too... any idea how to use the seed?
@@KLEEBZTECH
Do you mean using the same seed? If you uncheck the random box it allows you to keep using the same seed for testing.
will that work to keep the same character face once swapped?@@KLEEBZTECH
Genius asks. In Fooocus, put together an image of a woman with certain characteristics. Of all the ones he generated, one is the one I liked the most. I use Image Prompt and Face Swap with that photo so that they all come out with the same face, let's say so that it is the same person. It FUNCION. All the photos I order that way are the same person. But I see that by using face swap to be the same person doing different things, it limits a lot. Sometimes he uses the same tones from the original photo that he used in Face Swap, meaning if I ask for Bikini he makes it blue like Jean's jacket in the original photo and many things I ask for are limited by this original photo. If I don't use Image Prompt and ask to generate what you want you have much more freedom in creating. So the question is, can I generate without face swap to have all the freedom and then apply the face swap to that photo that I generated so that it is always the same person I want? That is, for example, put a 23-year-old blonde woman with green eyes, drinking coffee on that side, the photo is generated for me, but the face that was generated for me is nothing like the one I always have to use with Face Swap and that generated photo. give him the face I want? I don't know if I explain myself. I'll give you my Instagram so you can see the character that is always made with the same photo as I said with Image Prompt and Face Swap but this way I always get the same face or the same person but the results it gives me are greatly limited. @emilia_diaz_fit
So i need to know the best way to generare the same person doing different things in differents scenarios without being limited by the photo that i used to face swap. Is for a ig character it need to be the same person.
Thank you so much
Your Focus generates the pictures so damn fast, how did you do that? Mine needs 10 minutes to come up with 2 pictures..
Often I speed up generation in my videos but normally mine take about 30 seconds to 40 seconds depending on what I'm doing. I'm using a 3070 with 8 GB of vram. That is the biggest factor.
@@KLEEBZTECH still fast tho. I have Nvidia GTX 1660ti, 8GB as GPU. Do you know if that's slowing the process? I have Intel i7 CPU
Your video card is definitely the cause of it being slow. Video card is the most important thing.
@@KLEEBZTECH pfff, that's nasty.. Anyway, can you make a video or something about how to actually write a prompt to be followed? For example I'm creating a model and I keep writing that I want her with red eyes high contrast and nose piercing and I never get those. Maybe there are some tips and tricks idk to write as keywords for better results
I do plan on a video about prompting but will admit I do not consider myself great at that either. Still much of it is luck since prompt adherence is not great with these. They do say SD 3 will be much better. First thing I will mention is to disable styles to start with. They can sometimes cause issues. Masterpiece for example has the word earrings included. Also make use of inpainting if it doesn't get one or two things. Especially since it takes you so long to generate it can save an almost perfect image. Colors can be difficult since they can impact the wrong thing in the image. With inpainting you can tweak the prompt as you modify. Also be sure to use extra weight for the words or phrases you want to stress if needed. (nose piecing:1.3) for example. Hope those tips help if you were not already doing do them.
I'm having "Connection errored out" when i try face swap. What can i do?
Are you running locally or through something like colab? If the first time using it will download what is needed and if you have connection issues that might also be what is wrong. I am not sure.
What do you do when you need to put a face you have, on a face that's at an angle you don't have?
Sometimes if it is not a major difference it will work but otherwise FaceSwap will not easily work for that.
@KLEEBZTECH can you apply a face you already have, to the faces of a face grid?
The problem is that FaceSwap is going to try to make those faces turn towards the camera. You might be able to get it to do it a little but that is a challenge. I have not really experimented much with trying to get it to honestly.
do you have colab pro? because it tells me i dont have nvidia gpu. am i run out of free gpu credit? cause it gives error instead of producing the images.
I run it locally. I have never had good luck when trying on colab myself on the free teir.
@@KLEEBZTECH you are lucky to have strong pc. I am unfortunately not rich enough to get one though
fica desconectando
🙏🙏🙏🧡🧡🧡
Is a "style" basically a Lora?
Not really. It is just more text added to the prompt. ruclips.net/video/Sr2yXOSexy0/видео.html
It shows me to pay after few usage in fooocus is it free or not
Fooocus itself is free. I am not sure how you are using it so I really could not tell you much more without more information.
@@KLEEBZTECH I'm using it to replace outfits in the photos of my client but after a few usage it shows me to pay and I can't use GPU
You are not saying how you are using it. I am going to assume you are using it online somewhere? Are you using on Colab? If so then Colab only lets you use so much of their computer power each day for free. Then you have to wait until the next day or pay for the paid plan with Colab.
@@KLEEBZTECH yes I'm using it online on fooocus google collab
Is there any software for mac?
My understanding is that this will work on a Mac but not sure how well.
Can You train a model in fooocus?
No.
kohya_ss is what you're looking for
Does it do nfsw
Yes but you would probably be better off using checkpoint models and loras designed for that. I don't have a video yet on those.
I still can not figure out how to install stable diffusion smh.
Did you by chance watch my video on how to install? ruclips.net/video/j1WuQndmgFE/видео.html
stable fusion -> forge. all same and less simpler gui. we allready have all setup xD. only need correct model and prompt todo your image. yes if model is strained what you want. it create 0% its all trained
I know i'm on AMD with a 6900XT but how the fuck is it that fast :( takes 1minute20s for me to generate. EDIT : i'm a dumbass it's sped up.
😆 Yeah Lightning is pretty fast but mine takes about 35 seconds on a 3070 for 60 step quality.
@@KLEEBZTECH Oh... That was also a lightning one. Not using those for now.
Please someone give me a good GPU! 😭