If you find the video useful and would like to tip you can buy me some electricity for all these image generations. It is greatly appreciated! ko-fi.com/kleebztech
I've found this workflow useful for controlling the direction of the face: - Get a 3d model in pose - have a reference photo of your subject's face angled towards the direction you want - layering that over the 3d models face in canva - generating a new 3d model from the edited picture with "high stop at" pyracanny - using the both the new 3d model with face angle to generate the final picture + the orignal face reference picture This way you're able to have a mould of the faceshape of your subject, and you're able to use prompts and seeed images to insert a face with high consistency in the angle that you want and it only takes a couple of extra steps Hope this helps!
Great video! I have learned so much from your channel. Do you have any tips for creating consistent background locations for the characters? (Ex: bedroom with the same wall color, furniture, and layout in every generated image)
I do not.... But it is something I will think about and if I find a good way will pass along or make a video. I will have to play around with some ideas and see if I can come up with something.
This thing is so cool. Weirdly it's easier to manipulate the poses with the phone version than the mouse but even a quick test got some cool results. Thanks for the tip!
Enjoy! It is great for getting poses that are hard to prompt for. Sometimes a little more difficult with realistic images but really useful sometimes. Are all your images you use in your videos now done with Fooocus? Got some good music on your channel.
@@KLEEBZTECH thanks! I was using Bing/DALL-E but the restrictions on what it allows are so ridiculous that I've given up. I'm not talking nudity or anything even PG rated. Literally you have to fight it just to get a full length shot of somebody in a dress. I'm using it for facial models and then Face Swapping in Fooocus where needed (sadly I can't get it to do my existing models 100% accurate so I'm thinking there's a video about LoRA character training I should be watching 😆) And thanks. I'm glad you're enjoying the music.
I did recall you mentioning that about Bing so that was why I was curious if now doing in Fooocus for them. As for a good video on LoRA training... That I have not made myself since I have not figured it out enough to make a video yet. It is a slow process so hard to experiment.
@@KLEEBZTECH ah I thought I saw a LoRA video on your channel. I need to experiment with it myself some, for sure. It's interesting that Fooocus will generate the same character in a single session, like if you do 20 renders they'll usually be the same person. That's kind of cool. I need to figure out how to capitalize on that and maybe use it to figure out consistency.
Weird Rodney, I just get a picture of the mannequin with no background in the pose, not a person...why would that be? I'm following your settings. Edit: Ahh I hadn't turned off debug preprocessors :)
Not easily. Clothing is a challenge but if it is something simple then you might be able to get very close. I do have some other videos that cover different aspects of doing it with image prompts and such. Same person can be done mostly with Faceswap.
Hi bro. Nevered used fooocus. Though. Been using sd for long time. Always used controllnet. But it seems like case of fooocus is better work than sd controllnet. Like recognition.
I use several different tools but Fooocus is my primary since it can do most of what I need without much hassle. Sometimes I have to use others like ComfyUI or Forge but avoid when possible.
Thanks, awesome. I would pose myself by taking a video and taking screenshots. How's the fooocus works with changing a man to female or even kid in different poses? I've had some good results with holding a stick in my hand and saying it's a sword etc. :-)
Doesn't really matter much if the model is male or female, Fooocus will have no problem changing that to what you prompt. Really just the basic structure. And it sounds like Fooocus knows just as we did as kids that a stick makes a great sword!
Hey Rodney, first off thanks for the video it’s very helpful. I’m having an issue with trying to duplicate a character into a few different poses. It’s a baseball player that was generated with Fooocus. The main issue I’m having is the uniform style keeps changing. The colors stay correct (blue jersey and white pants) but they don’t stay the same as in my generated images. I’m using the main image as the image prompt and the pose image in the box next to it on canny. I even tried running the image prompt image at 1.0 on the slider. I’m using the same prompt as I used to generate the original image to generate my pose images. All the poses are coming out great but if I did get the character to stay consistent, that would be perfect and exactly what I need. Any suggestions? Oh, these are realistic images I’m trying to generate.
The most you can really do is increase the stop at and weight on the image prompt. Simple clothing is easier and inpainting is probably where you would need to try and correct. Would want to make sure to enable the image prompt and inpainting. Sadly no easy way. I do have some other videos on tips for more consistent characters but nothing that is a perfect solution.
If you find the video useful and would like to tip you can buy me some electricity for all these image generations. It is greatly appreciated! ko-fi.com/kleebztech
Am a heavy fooocus user and this has been helpful thank you!
You're welcome!
Great information and tutorial as always, thank you for all of your hard work!
My pleasure!
I've found this workflow useful for controlling the direction of the face:
- Get a 3d model in pose
- have a reference photo of your subject's face angled towards the direction you want
- layering that over the 3d models face in canva
- generating a new 3d model from the edited picture with "high stop at" pyracanny
- using the both the new 3d model with face angle to generate the final picture + the orignal face reference picture
This way you're able to have a mould of the faceshape of your subject, and you're able to use prompts and seeed images to insert a face with high consistency in the angle that you want and it only takes a couple of extra steps
Hope this helps!
Some good tips.
Great content as always! Thank you
Great video! I have learned so much from your channel. Do you have any tips for creating consistent background locations for the characters? (Ex: bedroom with the same wall color, furniture, and layout in every generated image)
I do not.... But it is something I will think about and if I find a good way will pass along or make a video. I will have to play around with some ideas and see if I can come up with something.
Thank teacher, ❤
You're welcome 😊
amazing
Very cool! Thanks!
Very useful thanks!
You're welcome!
This looks amazing. I'll be giving it a try later for sure!
This thing is so cool. Weirdly it's easier to manipulate the poses with the phone version than the mouse but even a quick test got some cool results. Thanks for the tip!
Enjoy! It is great for getting poses that are hard to prompt for. Sometimes a little more difficult with realistic images but really useful sometimes. Are all your images you use in your videos now done with Fooocus? Got some good music on your channel.
@@KLEEBZTECH thanks! I was using Bing/DALL-E but the restrictions on what it allows are so ridiculous that I've given up. I'm not talking nudity or anything even PG rated. Literally you have to fight it just to get a full length shot of somebody in a dress.
I'm using it for facial models and then Face Swapping in Fooocus where needed (sadly I can't get it to do my existing models 100% accurate so I'm thinking there's a video about LoRA character training I should be watching 😆)
And thanks. I'm glad you're enjoying the music.
I did recall you mentioning that about Bing so that was why I was curious if now doing in Fooocus for them. As for a good video on LoRA training... That I have not made myself since I have not figured it out enough to make a video yet. It is a slow process so hard to experiment.
@@KLEEBZTECH ah I thought I saw a LoRA video on your channel. I need to experiment with it myself some, for sure.
It's interesting that Fooocus will generate the same character in a single session, like if you do 20 renders they'll usually be the same person. That's kind of cool. I need to figure out how to capitalize on that and maybe use it to figure out consistency.
Perfect 👍👍👍
Thank you 👍
😍
Weird Rodney, I just get a picture of the mannequin with no background in the pose, not a person...why would that be? I'm following your settings.
Edit: Ahh I hadn't turned off debug preprocessors :)
That will do it. 😂
thanx😀😀😀
Welcome 😊
Is it possible to make same dress, same person, but different poses?
Not easily. Clothing is a challenge but if it is something simple then you might be able to get very close. I do have some other videos that cover different aspects of doing it with image prompts and such. Same person can be done mostly with Faceswap.
Bro para que salgan dos personajes como seria? osea peleando
Good job. I would like to ask, how can I control the direction of the lights?
In posemyart at the bottom left corner is the controls for that.
Hi bro. Nevered used fooocus. Though. Been using sd for long time. Always used controllnet. But it seems like case of fooocus is better work than sd controllnet. Like recognition.
I use several different tools but Fooocus is my primary since it can do most of what I need without much hassle. Sometimes I have to use others like ComfyUI or Forge but avoid when possible.
@@KLEEBZTECH thanks
Thanks, awesome. I would pose myself by taking a video and taking screenshots. How's the fooocus works with changing a man to female or even kid in different poses? I've had some good results with holding a stick in my hand and saying it's a sword etc. :-)
Doesn't really matter much if the model is male or female, Fooocus will have no problem changing that to what you prompt. Really just the basic structure. And it sounds like Fooocus knows just as we did as kids that a stick makes a great sword!
Hey Rodney, first off thanks for the video it’s very helpful. I’m having an issue with trying to duplicate a character into a few different poses. It’s a baseball player that was generated with Fooocus. The main issue I’m having is the uniform style keeps changing. The colors stay correct (blue jersey and white pants) but they don’t stay the same as in my generated images.
I’m using the main image as the image prompt and the pose image in the box next to it on canny. I even tried running the image prompt image at 1.0 on the slider. I’m using the same prompt as I used to generate the original image to generate my pose images. All the poses are coming out great but if I did get the character to stay consistent, that would be perfect and exactly what I need. Any suggestions? Oh, these are realistic images I’m trying to generate.
The most you can really do is increase the stop at and weight on the image prompt. Simple clothing is easier and inpainting is probably where you would need to try and correct. Would want to make sure to enable the image prompt and inpainting. Sadly no easy way. I do have some other videos on tips for more consistent characters but nothing that is a perfect solution.
@@KLEEBZTECH Ok thanks I’ll checkout your other videos
Does fooocus worjpk these days in colab?
I have used it just yesterday for an hour or so.
thanks video,i want to inpaint area to create girl,how to use pose like video
I'm working on a video that might help you with that. Hopefully should have it up tomorrow or the next day.
This might help. ua-cam.com/video/G5vC3Dl6kpU/v-deo.html
@@KLEEBZTECH thanks，i had learned,very cool,so i have a new question.please check the video discussion.
What's your computer specs? Mine takes forever to generate
i5 32GB RAM 3070 8GB but just installed a 4070 12GB.
does this thing have sexy poses with 2 characters if you know what i mean?
Probably depends on how "sexy". lol. But you can also adjust any of the included ones.
@@KLEEBZTECH thanks. Did you saw any, category or something like that? I just want to test stable diffusion censoring no shady business.
I would start with the search and try things like kiss.
focus dose not load, it keeps loding endless
What is it saying in the CMD window? Is it trying to download any missing models?
Nothing bro! Just watching ads!!
I would love to write bone prone Girl😂
I don't think UA-cam would like the results... 🤔