How does your avatar look when enhanced by AI?

[Nostoll]

[Istelathis]

I decided to give Dezgo  a try, there is an img2img option over there for any of you interested. These were upscaled using AnyRec

 

I'm wearing my kittycat as an avatar, wandering around SL so I am using her for these screenshots.

Original snapshot:

Dezgo img2img

 

Another snapshot

Dezgo

I had it set to very subtle changes, if I take it too high then everything starts to look wonky 

I prefer my own copy of easy diffusion, but I can't run it while playing around in SL so this is a good alternative for me, and best of all free 🙃

[Istelathis]

Here are a few I generated with InstantID 

The original image I used for most of these to generate with.

 

The results using a variety of prompts

 

The video I watched which introduced me to this was pretty interesting, here it is for those of you who want to try it locally from your computer (I have not installed it myself, I used the above link from huggingface)

 

Looking forward to anyone who uses this to share their results here  I just started playing around with it but have used up all of my GPU cycles on the site.  I'll probably install it when I get a new computer later on (hopefully 🤞)

Edited January 25 by Istelathis

[Istelathis]

Using the above picture in my previous post as the control image, but now using Dezgo, unfortunately mah eyes got crossed.

Lets see if I can fix that

..

I liked my face better in the first result.

 

Now trying out other features on the site, such as changing my hair color, removing the background, and inpainting myself in a different environment.

meh.. 

Yay a buffet!  

[Istelathis]

Another one done on InstantID

And below is another try with Dezgo's img2img using DreamShaper 8

I like how it changed a portion of my house to the cyberpunk theme I asked for on my avatar.  Ignore the extra fingers 😜

 

 

 

Edited January 31 by Istelathis

[AlexandriaBrangwin]

Which one is the Avatar?

[Cristiano Midnight]

Amor ❤️ (Gencraft / Firefly )

 

Edited February 2 by Cristiano Midnight

[AlexandriaBrangwin]

Oh no😆

[Istelathis]

After watching a new video regarding a new word being introduced to the web, "Vesperance" I was once again interested in goofing around with AI, and image generation.

I had an image created in Microsoft's copilot, which shows a lady at a campfire

Using the same picture I had in my previous posts to generate a new face for the pose which is below.

Below was the first result, which I liked.  I accidently left the template to watercolor though. I only used pose for the controlnet.

I removed the watercolor template, used canny and pose for the control net, bumping up the strength of pose to 0.9 and the below was the result.

I really like InstantID and look forward to getting my computer so I can run it locally.  For those interested in using it, you can play around with it at this site:

https://huggingface.co/spaces/InstantX/InstantID

It is really easy to play around with, and I would love to see some of the results everyone comes up with.  The main drawback of this site is, that you only get a few tries and it may take a while as it is free and other people are using the service.  For free image generation online, I just use bing.com, go into chat and prompt what I want drawn, it is also free.  

 

[Cristiano Midnight]

I went a bit crazy reimagining my avatar in Stable Diffusion

 

[MissSweetViolet]

Did this on https://playground.com/

Original image;

Ran it through with the following settings;

Seed
575652706

Guidance Scale
10

Sampler
Euler a

Model
Stable Diffusion XL

Initial Image Strength
80

Filter Style
RealVis XL

Refinement
25

Result;

[Ze Helix]

I have been playing with image2image on https://leonardo.ai/ using the free options. Mostly throwing in old screenshots some 10 years old or more.

 

[Cristiano Midnight]

A subtle image I did for an Instagram challenge in Gencraft, Stable Diffusion and Adobe Firefly.

 

[taisiyakarpenko]

[Wesley Spengler]

After seeing it mentioned a lot the past few days, I decided to try Leonardo myself. Interesting, but underwhelmed... Unlike Cristiano Midnight's results, it didn't exactly turn my avatar into the swarthy human cover guy I was hoping for... 🤣  But, fun for those of us who don't have the inclination to dig into the toolsets very deeply.

The three examples were from three different attempts at describing the desired result. 

[Claireschen Hesten]

I tried that Leonardo too with the pic i used the result wasn't too bad

[Wesley Spengler]

6 hours ago, Claireschen Hesten said:

I tried that Leonardo too with the pic i used the result wasn't too bad

Yeah, extremely cool result, actually.

[Clem Marques]

Before:

After:

[Cristiano Midnight]

23 hours ago, Wesley Spengler said:

After seeing it mentioned a lot the past few days, I decided to try Leonardo myself. Interesting, but underwhelmed... Unlike Cristiano Midnight's results, it didn't exactly turn my avatar into the swarthy human cover guy I was hoping for... 🤣  But, fun for those of us who don't have the inclination to dig into the toolsets very deeply.

The three examples were from three different attempts at describing the desired result. 

I am getting Clark Kent vibes on that last one though. You've got it in you. Good job on those, they look great. What did you use?

[Wesley Spengler]

9 hours ago, Cristiano Midnight said:

I am getting Clark Kent vibes on that last one though. You've got it in you. Good job on those, they look great. What did you use?

🤣 I didn't see that in the last one until you mentioned it! So true.

All were just plain Leonardo.ai. Unfamiliar with the tool, and limited to the free allocation of tokens, I was sort of guessing at what to put in the prompts or settings, apart from setting image-to-image mode as the blogs (New World Notes, etc.) have noted.

I've played some with Adobe's Firefly, which I know you've been incorporating too, and so far, the results aren't worthy of sharing. This is a lot of fun though, and you've gotten some particularly amazing results.

SL, our avatars, etc. have always sort of messed with my brain cells a bit (or a lot). Life-like AI renderings only dial it up more...

[Cristiano Midnight]

6 hours ago, Wesley Spengler said:

🤣 I didn't see that in the last one until you mentioned it! So true.

All were just plain Leonardo.ai. Unfamiliar with the tool, and limited to the free allocation of tokens, I was sort of guessing at what to put in the prompts or settings, apart from setting image-to-image mode as the blogs (New World Notes, etc.) have noted.

I've played some with Adobe's Firefly, which I know you've been incorporating too, and so far, the results aren't worthy of sharing. This is a lot of fun though, and you've gotten some particularly amazing results.

SL, our avatars, etc. have always sort of messed with my brain cells a bit (or a lot). Life-like AI renderings only dial it up more...

Feel free to contact me if you need any advice. I have started integrating AI into all aspects of my RL work (not just image generation), and I am fully immersed in it now (and going a bit nuts). Happy to help - this applies to anyone. Some people can be very proprietary about this stuff, which quite frankly, is stupid. We are all learning, it is a brave new world of our AI overlords.

[JeromFranzic]

Epic-realism model from tungsten.run. Original photo is in the Adult forum.

 

[ValKalAstra]

Doing some more experiments. For some stubborn reason only my broken mind can fathom, I decided to go the insane route. Make a character lora - then use SD to generate my SL character. Both the pictures are 100% entirely AI and yet could be mistaken for some touched up SL Flickr Shots, I feel.

Also heeeellooooooo uncanny valley!

[Cristiano Midnight]

Instagram challenges keep me occupied with AI lol:

(Gencraft, Adobe Firefly)

 

[ValKalAstra]

Does the world need more Nina Stickers? For sure! New LayerDiffusion is actually pretty neat. Having near on perfect transparency on generation is a big deal.