If I use the image prompt + text method, only the surroundings of the face will match the style, while the face itself will be directly transferred from the photo to the image without adopting the watercolor, clay, etc., style. How can I apply the style on the face as well?
I like your style. Good examples and didactic explanation. Let me ask a question: what is the best way to get a Full Body starting from photo with only the middle of the body?. Is it enough select PyraCanny with image?
Thanks! Glad you liked the tutorials of this channel. yes, In ImagePrompt section, upload the source image to use its face and select FaceSwap for it. Upload the full body image that you want to use its pose and select PyraCanny. Remember to increase stop at and weight like I said in this tutorial and don't forget to choose the right aspect ratio for a full body image result. For me, using PyraCanny was the best way. Test it and test other way if you needed.
When you run the default edition in Colab, the default model is juggernautXL_v8. If you want to use a special model instead, you need to install this model. I've created a video about installing the model in Colab. You can watch it to learn how to do it.
@@AIEdgeMastery i mean i don't want to install it i just want to replace the model i want to use with default version 8 without installing and i've asked other and told me to make copy in colab and then change according to need but i don't know how to do it
In Colab you can separate the last line, run the first 4 lines of code to clone fooocus, then click on Files icon on the left side. Open Fooocus folder, then open the models folder, then upload your file to the checkpoints folder. But I think it's proper for smaller files, models are pretty large files, and installing them is best.
Fooocus doesn't merge faces. This changes the source image based on your other input image. If you want to make the source image better and not change it completely, in inpaint use improve details and mask part of the face that you want to improve.
As far as I know Fooocus AI is primarily designed for desktop platforms and currently no official version for iPad. You can use it through a browser. To download and install it, ask your question on the github page, maybe they can help you.
Hi, is it possible that something has changed since the last update? I've noticed that "Mixing Image Prompt and Inpaint" no longer works. The result is largely the same as the original image. Does anyone know more?
Hi, I recently created a video about changing hairstyle, and in one section I used Mixing Image Prompt and Inpaint. It works and I did't see any problem. In the new version we have only some features. Your problem may be solved by changing the 'Weight' and 'Stop At' parameters. Don't mask extra part and disable the styles. Also, use prompt if needed.
I mostly use 13:19, 3:5, 4:7, 11:21, and 1:2. keep in mind that to get a better results, add 'full body' and something like shoe color in the text prompt.
But GPU disconnecting again and again in only 3 to 4 minutes, when we perform heavy task! Like when we add multiple images in image prompt. Android phone
When you use Colab, do your work step by step. For example, first do face swap, then change the pose and expression, then edit the clothes and hair, and then improve the details. In Colab's free plan, you have limited daily usage per account, and you can't do heavy tasks at the same time.
Google collab Gradio link isn't working from past 2 or 3 days, i'm using the same collab notebook to run the Fooocus but after all loading ,it doesn't give me gradio link, only giving a ip address at the end but that ip address doesn't connect, Kindly do reply me to solve this , Thanks
Test with a different account, and try a different web browser. This problem can be caused by a poor internet connection. Also, if you are using a VPN to use fooocus on Colab, I suggest you try with a different VPN.
Did you use a high quality image? What base model did you use? Disable the styles and try again. If you're using a high quality source image, you don't need to use upscaling.
@@AIEdgeMastery the default juggernaut model. i selected the "improve details..." option for inpainting instead of "inpainting/outpainting (auto)" and is working fine now.
If you upload your image in the image prompt section, increase the stop at and weight parameters. According to the video, if you upload two images in 'image prompt' and 'inpaint or outpaint', then you need to go to Developer debug mode, control tab, and active 'Mixing image prompt and inpaint'. Also if you use images in 'image prompt' and 'Upscale or Variation', then you need to go to Developer debug mode, control tab, and active 'Mixing image prompt and Vary/Upscale'.
The vary and upscale functions are used to modify and enhance the generated images. The vary function is used to create variations of an existing image generated by the AI model. When you use Vary, the model generates multiple images similar to the original image, but with minor differences. These differences can be in colors, textures, details, or other image elements. The purpose of this feature is to explore different possibilities and find the best variation of the generated image. The Upscale function enhances the resolution and image quality. This is especially useful if the original image was generated at a lower resolution.
Most of the time this problem may happen due to internet conditions. Check your internet and if you live in an area where you need to use a VPN, check the VPN status. If the problem is not related to these items, try using a different browser and a different user account. Do not forget that Colab has limitations for daily use with each account.
Thanks for the tips but I think it only works with people you don't know. If I try to insert a photo of myself the result is clearly not me, maybe a brother I don't know about, but not me ;-)
Thank you for your feedback! 🙂 You mentioned an interesting point. Faceswap technology, including that used in Fooocus AI, often works best with faces on which the AI has been extensively trained. Personal photos, especially if they're not part of the training data, can sometimes produce less accurate results.
Did you see the Minimal Requirement section on the github page? Based on that can you install fooocus on your computer? Give me more details of your problem.
When you get error, disconnect colab, check your internet connection, and try again. If it happens again, try a different browser. Don't forget that Colab has daily limits. If you do these things and get the error again, send me the error message, and more details like are you using a specific model? Do you want to generate new images, or use faceswap or....?
Thanks for your feedback. 😊 For this or any other tutorial, be sure to follow the tutorial first and do the same to learn how to use the ai tools, then test the different options. Sometimes you need to test several times and this does not happen only to you. Use these tools and don't give up.
Check your internet connection. If you should use a VPN in your location, try to stay connected to a single VPN server for the duration of your Colab session, also test with another one. Colab has a daily limit. Sometimes, the issue might be related to the browser. Test with another account and try using a different browser or clear your browser's cache. When creating an image, first disable additional settings such as styles and then add them to check the status in both modes. Sometimes, applying several styles can cause problems.
Note that you must use the same model that you used to create the image for this part. The image used must be clear. Change the stop at and weight values according to the video. Create several images and you will find similar items among them.
@@AIEdgeMastery Yes, I did use Face Swap. But the result looks like western. I know it is not 100% as same as the original face, but somehow the face becomes western. If a type "asian" into the prompt, it may a little bit better.
I didn't test the face swap with an Asian face. When you create an image and do not specify the nationality, it is normal that the output will be based on the default specified for that tool. To get the desired output, the nationality must be mentioned in the prompt. You can also get help from Loras on the Civit ai website.
BEST tutorial everrrrr!! ty
Thank you so much! Glad it helped! 🙌😊
Straight to the point!
Thank you for the video, really helpful. ❤
Glad you liked it! 😊
Спасибо!
Very useful, simple tutorial video!! thank you!!
Glad it was helpful!
"Wow, this was so informative! Thanks for sharing!"
Glad it was helpful! 😊
Excelent, really excelet tutorial! Complete on the point and explaining all the functionalities!!!
Glad it was helpful! Thanks for your kind feedback! 😊
nive tutorial.I hope you upload videos about fooocus lab more often
Thanks. I will make more videos about this. 🙂
If I use the image prompt + text method, only the surroundings of the face will match the style, while the face itself will be directly transferred from the photo to the image without adopting the watercolor, clay, etc., style. How can I apply the style on the face as well?
I like your style. Good examples and didactic explanation. Let me ask a question: what is the best way to get a Full Body starting from photo with only the middle of the body?. Is it enough select PyraCanny with image?
Thanks! Glad you liked the tutorials of this channel.
yes, In ImagePrompt section, upload the source image to use its face and select FaceSwap for it. Upload the full body image that you want to use its pose and select PyraCanny. Remember to increase stop at and weight like I said in this tutorial and don't forget to choose the right aspect ratio for a full body image result.
For me, using PyraCanny was the best way. Test it and test other way if you needed.
to get a full body , just put shoes in the prompt.
very useful video,Thank you
Glad it was helpful! 😊
thank you. it is very useful.
Glad to hear that. 😊
hey do you have any idea on how can i make juggernaut x model as a default model in fooocus?
Do you have fooocus installed locally on your computer or are you using it in Colab?
@@AIEdgeMastery i'm using it on colab and also in kaggle
When you run the default edition in Colab, the default model is juggernautXL_v8. If you want to use a special model instead, you need to install this model. I've created a video about installing the model in Colab. You can watch it to learn how to do it.
@@AIEdgeMastery i mean i don't want to install it i just want to replace the model i want to use with default version 8 without installing and i've asked other and told me to make copy in colab and then change according to need but i don't know how to do it
In Colab you can separate the last line, run the first 4 lines of code to clone fooocus, then click on Files icon on the left side.
Open Fooocus folder, then open the models folder, then upload your file to the checkpoints folder.
But I think it's proper for smaller files, models are pretty large files, and installing them is best.
Nice share
Glad you liked it! 😊
thank you 🥰🥰🥰🥰🥰
You’re welcome 😊
Can you indicate a face swap like the one you had on faceapp? That mixes two faces and creates another, not just changing the face.
Fooocus doesn't merge faces. This changes the source image based on your other input image. If you want to make the source image better and not change it completely, in inpaint use improve details and mask part of the face that you want to improve.
Can we download it on iPad Pro M1
As far as I know Fooocus AI is primarily designed for desktop platforms and currently no official version for iPad. You can use it through a browser. To download and install it, ask your question on the github page, maybe they can help you.
is there any way to auto select face in inpaint ? so that you dont have to keep painting the face?
As far as I know, you have to do this manually.
@@AIEdgeMastery How about auto-segmentation? can you use it foocus?
Hi, is it possible that something has changed since the last update? I've noticed that "Mixing Image Prompt and Inpaint" no longer works. The result is largely the same as the original image. Does anyone know more?
Hi, I recently created a video about changing hairstyle, and in one section I used Mixing Image Prompt and Inpaint. It works and I did't see any problem. In the new version we have only some features. Your problem may be solved by changing the 'Weight' and 'Stop At' parameters. Don't mask extra part and disable the styles. Also, use prompt if needed.
What is the aspect ratio to get a full body image?
I mostly use 13:19, 3:5, 4:7, 11:21, and 1:2.
keep in mind that to get a better results, add 'full body' and something like shoe color in the text prompt.
But GPU disconnecting again and again in only 3 to 4 minutes, when we perform heavy task! Like when we add multiple images in image prompt. Android phone
When you use Colab, do your work step by step. For example, first do face swap, then change the pose and expression, then edit the clothes and hair, and then improve the details.
In Colab's free plan, you have limited daily usage per account, and you can't do heavy tasks at the same time.
Google collab Gradio link isn't working from past 2 or 3 days, i'm using the same collab notebook to run the Fooocus but after all loading ,it doesn't give me gradio link, only giving a ip address at the end but that ip address doesn't connect, Kindly do reply me to solve this , Thanks
Test with a different account, and try a different web browser.
This problem can be caused by a poor internet connection. Also, if you are using a VPN to use fooocus on Colab, I suggest you try with a different VPN.
Gradio Live is down for quite some time.
Im using the first method and the swapped face is very blurry. Tried upscaling as well. Any ideas?
Did you use a high quality image? What base model did you use?
Disable the styles and try again. If you're using a high quality source image, you don't need to use upscaling.
@@AIEdgeMastery the default juggernaut model. i selected the "improve details..." option for inpainting instead of "inpainting/outpainting (auto)" and is working fine now.
When you use faceswap, you should first use inpaint or outpaint method, then change the method if you want to improve the special part.
what is source image?????? it doesn't use the picture I uploaded as a reference at all
If you upload your image in the image prompt section, increase the stop at and weight parameters.
According to the video, if you upload two images in 'image prompt' and 'inpaint or outpaint', then you need to go to Developer debug mode, control tab, and active 'Mixing image prompt and inpaint'.
Also if you use images in 'image prompt' and 'Upscale or Variation', then you need to go to Developer debug mode, control tab, and active 'Mixing image prompt and Vary/Upscale'.
@@AIEdgeMastery thank you, I don't understand the part with vary/upscale. what's the function behind it
The vary and upscale functions are used to modify and enhance the generated images.
The vary function is used to create variations of an existing image generated by the AI model. When you use Vary, the model generates multiple images similar to the original image, but with minor differences. These differences can be in colors, textures, details, or other image elements. The purpose of this feature is to explore different possibilities and find the best variation of the generated image.
The Upscale function enhances the resolution and image quality. This is especially useful if the original image was generated at a lower resolution.
@@AIEdgeMastery thank you!
The colab is dissconnecting when I try to use it :(
Most of the time this problem may happen due to internet conditions. Check your internet and if you live in an area where you need to use a VPN, check the VPN status. If the problem is not related to these items, try using a different browser and a different user account. Do not forget that Colab has limitations for daily use with each account.
Thanks for the tips but I think it only works with people you don't know. If I try to insert a photo of myself the result is clearly not me, maybe a brother I don't know about, but not me ;-)
Thank you for your feedback! 🙂
You mentioned an interesting point. Faceswap technology, including that used in Fooocus AI, often works best with faces on which the AI has been extensively trained. Personal photos, especially if they're not part of the training data, can sometimes produce less accurate results.
@@AIEdgeMastery Thank you for the answer! Maybe personal trained Loras are still necessary for this then 😞
Use real images for faceswap does not really work, right? i get superbad results
You may not get perfect results, but you can use a realistic base model to get better results.
fooocus this is my computer i couldn't install please help me😢
Did you see the Minimal Requirement section on the github page? Based on that can you install fooocus on your computer? Give me more details of your problem.
after press generate button fooocus ai give me error in web ui
Do you use Colab?
@@AIEdgeMastery yes
When you get error, disconnect colab, check your internet connection, and try again.
If it happens again, try a different browser. Don't forget that Colab has daily limits.
If you do these things and get the error again, send me the error message, and more details like are you using a specific model? Do you want to generate new images, or use faceswap or....?
Hmm... Why the face is still digferent
At what time do you see different faces in the video? 🤔
Very good video❤ but nothing from this or other videos works for me...🤮 Like the ai has a mind of its own and doesnt like my face😂
Thanks for your feedback. 😊
For this or any other tutorial, be sure to follow the tutorial first and do the same to learn how to use the ai tools, then test the different options. Sometimes you need to test several times and this does not happen only to you. Use these tools and don't give up.
Does this still working?
If you have a problem, they may have made a change in the new updates. Did you test with the images you made with fooocus?
i always get connection lost
When will your connection be disconnected? When you generate images with text prompt, or when you use face swap, or others?
@@AIEdgeMastery both times Sir
Check your internet connection.
If you should use a VPN in your location, try to stay connected to a single VPN server for the duration of your Colab session, also test with another one.
Colab has a daily limit. Sometimes, the issue might be related to the browser. Test with another account and try using a different browser or clear your browser's cache.
When creating an image, first disable additional settings such as styles and then add them to check the status in both modes. Sometimes, applying several styles can cause problems.
@@AIEdgeMastery okay sir I will try this but I am using 5G internet and stable VPN connection
wow
Its not working
Does that mean you can't use Colab, or these ways don't work for you?
Do you try all the ways?
Give me more details to help you.
@@AIEdgeMastery the face is not same its different everytime even I am using the image this created
Note that you must use the same model that you used to create the image for this part. The image used must be clear. Change the stop at and weight values according to the video. Create several images and you will find similar items among them.
very bad result...
It doesn't work for Asian face.
Did you create image of an Asian face with fooocus and then try to use Face Swap?
@@AIEdgeMastery Yes, I did use Face Swap. But the result looks like western. I know it is not 100% as same as the original face, but somehow the face becomes western. If a type "asian" into the prompt, it may a little bit better.
I didn't test the face swap with an Asian face. When you create an image and do not specify the nationality, it is normal that the output will be based on the default specified for that tool. To get the desired output, the nationality must be mentioned in the prompt. You can also get help from Loras on the Civit ai website.
Running locally; keep getting error about -src is not a numpy array, neither a scalar
When does this error occur? Did you upload the image and then this happened?