Jump to content

How does your avatar look when enhanced by AI?


Danielle Atheria
 Share

Recommended Posts

merged03.thumb.jpg.0eecdee1edf56e27442e36a7e3c5238d.jpg

 

I had a bit of fun playing around with this one, I initially generated an image to image using concept art as part of the prompt, then I created the background as a separate image, used GIMP to put them together, and then played around with the colors.  Not all that great, but it was fun, I think eventually I may make a hobby of this and hopefully get better at it with time.

The avatar is based from this image I had used earlier for the forum

Snapshot_467.thumb.png.5a4566303b659ccbe1f950fc828f85ee.png

The background was just trial and error using various prompts until I generated one I enjoyed enough to use.  Don't mind the multiple goofups in the final picture, an artist I certainly am not.

Edited by Istelathis
  • Like 7
Link to comment
Share on other sites

Goofing around again, I wanted something old school, a little retro nostalgia

Making a background from this picture in Winter Wonderland

scaleddown.thumb.png.dcc61b52eab7e0b7e955ee6a8e842915.png

And using my avatar from this one

Snapshot_558.thumb.png.c033f71aa272958ef3c82b682ff60b69.png

and changing each of them to to much lower quality, and then combining them in GIMP

16bitimg2x.thumb.jpg.830ca651a89a38d7004ee8fe66029b58.jpg

What have I done¿  It kind of reminds me of Daggerfall which I'll include a picture below.  I shrunk my avatar, so the pixels aren't as noticeable though :( I needs my pixels!

5891186491607510269gol1.jpg

Daggerfall screenshot

 

Edit: 

Cheating a bit by adding pixels from GIMP to my pixie, a pixelized pixie.

16bitpix2x.thumb.jpg.75ac2a6a06d0e72fb2d7ae5feca03991.jpg

Edited by Istelathis
  • Like 6
Link to comment
Share on other sites

I still haven't created a LORA, I just have too many avatars with different faces to create one with.  I prefer instead to try my best to use an existing image and try my best to convert it as close as possible to the original.  Here is one such attempt, but it doesn't quite match up right.

Snapshot_102.thumb.png.4cc5c4537b554ce88ccdb24a85ae6e3a.png

aniFina2l.thumb.jpg.e17e75e06f25af48dfc8080deaa4a0ae.jpg

 

I think my avatar's lips are too puckered for the configuration I have setup.  I have too much makeup on as well, I have to find a different skin, shape my lips a bit different, and I might have better luck.  I've had more success in previous posts on this thread.  Still, I am pretty impressed with the overall pose of the image, the lines that easy diffusion tries to stick true to.

  • Like 6
Link to comment
Share on other sites

16 hours ago, Istelathis said:

I still haven't created a LORA, I just have too many avatars with different faces to create one with.  I prefer instead to try my best to use an existing image and try my best to convert it as close as possible to the original.  Here is one such attempt, but it doesn't quite match up right.

Snapshot_102.thumb.png.4cc5c4537b554ce88ccdb24a85ae6e3a.png

aniFina2l.thumb.jpg.e17e75e06f25af48dfc8080deaa4a0ae.jpg

 

I think my avatar's lips are too puckered for the configuration I have setup.  I have too much makeup on as well, I have to find a different skin, shape my lips a bit different, and I might have better luck.  I've had more success in previous posts on this thread.  Still, I am pretty impressed with the overall pose of the image, the lines that easy diffusion tries to stick true to.

A big part of the difference is the lighting. The ambient light in the first image is way more than in the second. Plus there is a color cast. I suggest you consider tweaking on the EEP/Windlight settings before worrying too much about a skin.

  • Like 4
Link to comment
Share on other sites

56 minutes ago, Leora Greenwood said:

I enjoy your finished images and would also value seeing the image with which you began your journey.

Alright :) To get a consistent look, all of my AI images are based on a headshot of my avatar:

headshot.thumb.jpg.fe81a1e75f858188278cbe3dd645059d.jpg

 

Gencraft has tech that is similar to ControlNet Canny for Stable Diffusion that will take elements of the source image and use it in the generated results.

For the ""bad hair day picture", it generated this from my headshot:

aleanathletic56e65cd8-eeb0-457a-b61c-eba88188e29d(1).thumb.png.459d203eee28904e766bfd0f828f8334.png

From there, I cleaned it up and then expanded it in Photoshop:

Honeyview_aleanathletic56e65cd8-eeb0-457a-b61c-eba88188e29d.thumb.png.5f71f37cf1c1dbf2c5e2cc63c3ae3f89.png

Then I rotated and filtered the final image a bit to make it more visually interesting.

53426740005_b9ffbe16cb_k.thumb.jpg.d9e3bb2524e0af047ef42577fabf76a1.jpg

  • Like 8
  • Thanks 1
Link to comment
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
 Share

×
×
  • Create New...