How to fix AI art faces : Midjourney, Stable Diffusion, Dalle 2
ฝัง
- เผยแพร่เมื่อ 28 ก.ค. 2024
- In this video, I show you how to fix facial issues with your A.I.-generated images.
The first two methods use free photo restoration apps and the third is a subscription-based graphical program.
I take you through each method step by step, so you can select the most suitable method that works for you.
ARC by Tencent:
arc.tencent.com/en/ai-demos/f...
GFP GAN (You just need a google account & sign in) :
colab.research.google.com/dri...
Photoshop:
www.adobe.com
I am not paid, or affiliated with any of these companies, I just make these videos for fun and to share the information I have found.
Socials:
/ promptmuse
Blog:
www.promptmuse.com - แนวปฏิบัติและการใช้ชีวิต
![บุษบา - เมนทอล [Official MV]](http://i.ytimg.com/vi/IGnWPokSEis/mqdefault.jpg)
Don't forget code former! It keeps more of the texture and even adds some in.
I just tried it and it's awesome! Thank you!
All good methods! Depending on if it was only the eyes that needed "fixing", I usually just fix the eyes and then mask them in photoshop with the original, so it's the only bit restored.
Loving this new series ❤️
Thanks !
Always worth mentioning the free image editor GIMP if you don't want to pay for Photoshop.
Good shout 👍
Cool been going through a lot of your vids since subbing. Been using midjourney for our livestream Roleplaying sessions (mostly Sci-fi). AI has been a great help in conjuring up cool things without paying a ton of money for our hobby. I use midjourney so much the subscription is well worth it.
Thanks for posting. Tip 1 works a treat...
"to the instagrams"
I love her
This is one of my favourite tools. Note I have found even if you downscale most AI art it usually re upscales fine a limitation of this tool I have found it only accepts 2000 px by 200 px maximum
Great video! I see that you duplicated the eyes in the photoshop section, I would suggest after duplicating the eyes that you change the direction of the spotlight. Hope that makes sense. Thanks for a great video ✌️
Great tip, thank you !
Its interesting that the 'face correction' option sometimes doesent work, thankfully, tencent fix it then.
Thanks lady.
What do you do for a living? You seem to have so many varied talents (photoshop, ai, unreal engine, programming, etc)
How do I fix the eyes?
I did some of these edits, but the next time I went back to Stable Diffusion it asked me, "Why have you been screwing around with my artwork?"
Our overlords know EVERTHANG
Thank you for a great video, i've been trying out arc tencent, which on the whole is really good, but it completely removes blue eyes, is there something that can fix that, do you know of a program that can restore the blue eyes? I tried lunapic which was terrible.
Hi, great content. Can you show how to create the „magic avatars“ which are popular with Lensa App, on your own, using free ai tools?
Ooo Yes this would be cool to show, thanks for the idea. `I’ll look into it for sure
does it work if clients want their required detail fashion dresses along their faces in illustration.. how can we do with these AI tools
Thank you
You're welcome 🫶
Is there a way to use GFP Gan to restore and reshape the entire photo, rather than merely the face?
Not yet, but I’m sure there will be one soon. Or perhaps one I’m currently unaware of :)
Have you tried a male face for comparison on GFP GAN? Seems to me that mirroring / harmonising a face makes it look more feminine.
I have not, will give this a go. Thanks for the suggestion 👍
Great! Why are they showing us all the code? Can you copy and paste it somewhere?
cant get arc to work...keps saying Failed to return results. Please try again later
At 2:46 " user interface is showing you the executions being run but don't worry about that you don't need to know" Yeah thats what ai is going to tell us " dont worry about all this code".. "just click on accept".. Shouldnt we know what its doing.. so it doesnt have bad intentions?
At the moment the code is open source, so those with the means to learn how to add functionality can do so.
Its still free and open source.
But we are paying for the fancy UI, filters and added features etc you do have to pay for that to be developed.
Stable diffusion which is another ai art generator is free if you can run it locally on your computer (have enough GPU) otherwise you have to pay for a google collab pro account.
But I fear we have larger concerns in the world than being able to generate ai art for free in my POV but I do understand ppls concerns as well as it’s developing at a rapid pace.
Companies will use it instead of hiring a room of artists. I know this from experience.
Great comment though, it’s good to hear everyone’s opinions on ai.
Hey if you ever see this, can you point me to the image in the thumbnail?
Need one to fix hands now.
Haha a work around is to add “anatomically correct hands” to the prompt. It kinda works :)
@@promptmuse Nice one to try. Thank you!!
FIRST TWO didn't work, multiple faces in a group and nothing was changed, the second link didn't work but when i searched and found a few options on it, it still didn't fix MULTIPLE faces that have gone wonky with DALL-e. there has to be a way as i'm seeing TONS of these images used...
very nice! i planed to use this for my game-cards!
Amazing ! Do you have a instagram for your work/ progress?
@@promptmuse uhm currently not but i think i need to create one
I've been using faceapp on my iphone.
"restore face" dont work for me in make a bug in "AI1111 stable diffusion"
Great, AI + CCP Tencent... what could possibly go wrong!?
“May you live in interesting times”
@@promptmuse don't curse me! Haha
Well-produced videos. Keep up the good work.
@@rokko_hates_japan Haha I think at this rate we are all cursed 😂 Cheers !
I got this from another video, but it’s easier to fix in ps by opening your image, filter / neural filters/ smart portrait / eye direction. Just move the slider one or two stops and it will start working. No wonky eyes!
American media keeps telling us to worry about China, but the rest of the world is more worried about the US.
Yep, it makes sense to finish with ps. Plus that way we can take some credit for the art rather than just for the good prompt. 🤭
This is true!
2 minutes and the rest
wtf draw 4 the photoshop one a glasses 🤓
Impossible your photo to be junky !
Gfpgan very bad idea. Identifity zerolless!