Hey guys, I found a prompt that is satisfactory for me and that's selecting the hands and for the prompt, write "Jewelry", it usually fixes the defects. But I like the tips in this channel as well!
Thanks for your feedback and sharing this experience! Which inpainting method did you use? Inpaint or outpaint, or Improve detail? Did you use styles or disable them?
Glad you liked it! To use fooocus, see the minimal requirement section on the github page for running Fooocus locally. If your device's capability are lower than what they say, you may not be able to use Fooocus locally. So you can use it online and run it in Google Colab. In colab free plan, you have limited daily usage for each account and it is reset every day and you can use it again. On fooocus github page, which is in the description, you can see Colab section. Click on Open in Colab and run it from there. I have a playlist about Fooocus, if you don't know how to use it in colab, you can watch the first part of the first video which is about Creating Consistent Characters to see how I run it on Colab.
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.
Thanks for your suggestion! While the stop hand gesture can help in some cases, it may not always be the best choice for achieving natural looking hands in AI generated images. Different gestures can convey different emotions and contexts, which may work better depending on the scene.
Hi there, you probably has answered or has a video about it, i am pretty new with all of this. I see in your vide your SD has something with "styles", "aspect ratio" and the outpaint, i don't have such options in mine, they seem pretty cool tho, have you modified it? it's a different webui? If you have a video let me know!
Hi. If you've never used fooocus AI before and need more information on using face swapping, changing clothes and more, check out my stable diffusion fooocus playlist. For settings you can't see, when the interface is running, you can see 'Input Image' and 'Advanced' options below the prompt box. Check them out and you can see more options.
No, when I worked with the Inpaint tool I used the default settings and did not use the model and Lora that I downloaded in the previous section. also, I didn't use styles but it's optional and you can use styles based on your needs. Test with and without it.
Try using a different browser. If you still have problems, there are two ways. First, try to download the file locally. Then upload it directly from your desktop to Colab. Second, search for using civit ai API key.
As far as I know, you have to buy an account for 4x upscaling. Use Krea AI and Leonardo AI and test them. You may find them suitable for your work. I have already made a video comparing these two tools.
Hi. I deployed the fooocus fast API on runpod to try different pose or faceswap etc... The results are different interms of image quality and blending quality. It's generating good results on notebook but not when deployed its API on runpod...why? Plz, let me know.
@@AIEdgeMastery I used fooocus fast API and th settings are same...also images are same but very different weird artifacts...let me know your email so that I send you results...Thanks
I cannot help with the reason for the difference in results. If there is another case, ask. The email is available in the About section of the channel.
I agree with you. This tool is very useful and has good features, but it still has room for improvement and can become a great tool with effective updates.
I have written the same code as your lora code for colab, but it always errors and colab says the file is corrupted. Any suggestions? Or maybe you can make a tutorial to upload lora files in colab quickly? Hehe
Copy the code from the description. Make sure that your internet connection is fine, test it again at different times, and let me know if it works for you or not.
Hey guys, I found a prompt that is satisfactory for me and that's selecting the hands and for the prompt, write "Jewelry", it usually fixes the defects. But I like the tips in this channel as well!
Thanks for your feedback and sharing this experience! Which inpainting method did you use? Inpaint or outpaint, or Improve detail?
Did you use styles or disable them?
E
I hope SD3 will be better with hands.
Ilike the most the 3rd one method!! thank you again!
You're welcome! 😊
Thanks for tutorial. For Fooocus Inpaint, it need to be done by installing it locally? No web version?
Glad you liked it! To use fooocus, see the minimal requirement section on the github page for running Fooocus locally. If your device's capability are lower than what they say, you may not be able to use Fooocus locally. So you can use it online and run it in Google Colab.
In colab free plan, you have limited daily usage for each account and it is reset every day and you can use it again.
On fooocus github page, which is in the description, you can see Colab section. Click on Open in Colab and run it from there.
I have a playlist about Fooocus, if you don't know how to use it in colab, you can watch the first part of the first video which is about Creating Consistent Characters to see how I run it on Colab.
COOL Thanks
Glad You liked it!
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.
Try making the person to do the stop hand gesture to properly fix the hands problem
Thanks for your suggestion! While the stop hand gesture can help in some cases, it may not always be the best choice for achieving natural looking hands in AI generated images. Different gestures can convey different emotions and contexts, which may work better depending on the scene.
Hi there, you probably has answered or has a video about it, i am pretty new with all of this. I see in your vide your SD has something with "styles", "aspect ratio" and the outpaint, i don't have such options in mine, they seem pretty cool tho, have you modified it? it's a different webui? If you have a video let me know!
Hi. If you've never used fooocus AI before and need more information on using face swapping, changing clothes and more, check out my stable diffusion fooocus playlist.
For settings you can't see, when the interface is running, you can see 'Input Image' and 'Advanced' options below the prompt box. Check them out and you can see more options.
did you've used lora while using inpainting or styles? le me know
No, when I worked with the Inpaint tool I used the default settings and did not use the model and Lora that I downloaded in the previous section.
also, I didn't use styles but it's optional and you can use styles based on your needs. Test with and without it.
I can download the Realistic Stock Photo in the colab it's have error Un Authoring Username. how to fixed it?
Try using a different browser. If you still have problems, there are two ways. First, try to download the file locally. Then upload it directly from your desktop to Colab. Second, search for using civit ai API key.
Can you recommend free website to upscale image 4x without loosing quality?
As far as I know, you have to buy an account for 4x upscaling. Use Krea AI and Leonardo AI and test them. You may find them suitable for your work. I have already made a video comparing these two tools.
Hi. I deployed the fooocus fast API on runpod to try different pose or faceswap etc... The results are different interms of image quality and blending quality. It's generating good results on notebook but not when deployed its API on runpod...why? Plz, let me know.
Hi. I haven't used the runpod yet. Is your Notbook GPU the same as the one you used in runpod? Are you testing with the same image and settings?
@@AIEdgeMastery I used fooocus fast API and th settings are same...also images are same but very different weird artifacts...let me know your email so that I send you results...Thanks
I cannot help with the reason for the difference in results. If there is another case, ask. The email is available in the About section of the channel.
Overall, the inpaiting model is effective, but it still needs to be improved.
I agree with you. This tool is very useful and has good features, but it still has room for improvement and can become a great tool with effective updates.
Yoou can create picture in Focus and fix the hands with Flux inpaint. Works like magic.
Yes flux inpaint works well. There are different options and it depends on which one works best for you to use.
I have written the same code as your lora code for colab, but it always errors and colab says the file is corrupted. Any suggestions? Or maybe you can make a tutorial to upload lora files in colab quickly? Hehe
Copy the code from the description. Make sure that your internet connection is fine, test it again at different times, and let me know if it works for you or not.
@@AIEdgeMastery okay😁👌
For ,me the best results was using METHOD: Improve Detail (face hand, eye, etc) + PROMPT.
It depends on your image. Based on your image, you may need to choose a different method. 👌
*AND THE LINK TO THE COLAB TO COPY IT ?*
All you need is in the description.
is there any for anime style?
This Lora is also used for Anime style. Try all these works that I did, for Anime style, and let me know which one can change your image the best.
Thank you for your video
civit link not working
Please guide
You're welcome! I updated the Lora link, model link, and Colab codes in the description.