• There is NO official Otland's Discord server and NO official Otland's server list. The Otland's Staff does not manage any Discord server or server list. Moderators or administrator of any Discord server or server lists have NO connection to the Otland's Staff. Do not get scammed!

128x128 AI enhanced sprites

We live in different times than 6 years ago
Everything is moving forward technology
graphics card, etc.
you don't have to have several million thousands of data
to create something interesting
AI now generates images like it used to
have to put a lot of work into it, this is very sad but true
 
Sprites 4043-4078-3.pngMy newest results achieved from experimenting with different models, here I have blended multiple models to produce this output.
 

Attachments

  • Sprites 4331-4366-3.png
    Sprites 4331-4366-3.png
    1.4 MB · Views: 202 · VirusTotal
  • Sprites 245076-245111-3.png
    Sprites 245076-245111-3.png
    1.2 MB · Views: 220 · VirusTotal
  • Sprites 0-143-0.png
    Sprites 0-143-0.png
    1.5 MB · Views: 202 · VirusTotal
Love the sprites, but i think i have fell in love with chainner, it's absolutely amazing :D thank you so much for being so open about the process and apps you use, chainner has legitimately improved my life
 
Love the sprites, but i think i have fell in love with chainner, it's absolutely amazing :D thank you so much for being so open about the process and apps you use, chainner has legitimately improved my life
Right bro? You are welcome! This thing is absolutely amazing! It has to be the best unknown program around right now, and that is only because of how early it is in development right now, but those guys really put out some updates!

Better believe it's the tool I have used to work around the issue of the upscaled sprites overlapping :D
 
Ok so I have started dabbling with trying to "enhance" sprites before making them into a dataset. I will share here with you guys some of my results so far, and hopefully someone can give me their opinion on what they think. I will include the images in pairs, the ones which are "tuned" have "tuned" in the filename, but I will try to make sure they are always on the right side. You should see more dark lines and shadows on the "tuned" versions. I ran a few samples. Please let me know what you guys think

Sprites 104373-104408-3_5.pngSprites 104373-104408-3_5-tuned.png
Post automatically merged:

Sprites 108405-108440-3_3.pngSprites 108405-108440-3_3-tuned.png
Post automatically merged:

Sprites 104157-104192-3_7.pngSprites 104157-104192-3_7-tuned.png
Post automatically merged:

Sprites 111141-111176-3_6.pngSprites 111141-111176-3_6-tuned.png
 

Attachments

  • Sprites 104157-104192-3_7.png
    Sprites 104157-104192-3_7.png
    67.5 KB · Views: 7 · VirusTotal
Love the sprites, but i think i have fell in love with chainner, it's absolutely amazing :D thank you so much for being so open about the process and apps you use, chainner has legitimately improved my life
Some models you find in
  • Automatic11111
  • Chainner
  • ComfyUI

You can run directly some models under upscayl.

People say RealSRGAN is the best upscaler up until today, from my tests this is not true for like 90% of the test data.
There's plenty places where you can compare models and finetuning of settings, even people mixing models inside a workflow.

You can find benchmarks in the websites below:
You can also find some further papers here: https://awesomeopensource.com/search?q=upscaling

Nowadays SD (StableDiffusion) models are becoming more popular but there's plenty other techniques. I'd start with the models below because they are all awesome:


CCSR alone is absolutely wonderful, for most of you running it through fal.ai should already be enough.
Post automatically merged:

Any tutorial?

Depends what you want to use, if it's Chainner just follow the readme of the repo below:

For ComfyUI you can just follow the instructions here: GitHub - comfyanonymous/ComfyUI: The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. (https://github.com/comfyanonymous/ComfyUI)
Just a disclaimer that you should have a NVIDIA GPU. I could not use it in AMD GPU thoughts, I tried even rewritting some libraries to avoid CUDA but couldn't hahaha some guys showed a tutorial in directML but I've tried too and models too big don't run as they cannot allocate the torch vectors in GPU.

Alternatively you can run old models with their GUI version, there's plenty that are free all over github (e.g: Releases · AaronFeng753/Waifu2x-Extension-GUI (https://github.com/AaronFeng753/Waifu2x-Extension-GUI/releases))
 
Last edited:
Some models you find in
  • Automatic11111
  • Chainner
  • ComfyUI

You can run directly some models under upscayl.

People say RealSRGAN is the best upscaler up until today, from my tests this is not true for like 90% of the test data.
There's plenty places where you can compare models and finetuning of settings, even people mixing models inside a workflow.

You can find benchmarks in the websites below:
You can also find some further papers here: https://awesomeopensource.com/search?q=upscaling

Nowadays SD (StableDiffusion) models are becoming more popular but there's plenty other techniques. I'd start with the models below because they are all awesome:


CCSR alone is absolutely wonderful, for most of you running it through fal.ai should already be enough.
Post automatically merged:



Depends what you want to use, if it's Chainner just follow the readme of the repo below:

For ComfyUI you can just follow the instructions here: GitHub - comfyanonymous/ComfyUI: The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. (https://github.com/comfyanonymous/ComfyUI)
Just a disclaimer that you should have a NVIDIA GPU. I could not use it in AMD GPU thoughts, I tried even rewritting some libraries to avoid CUDA but couldn't hahaha some guys showed a tutorial in directML but I've tried too and models too big don't run as they cannot allocate the torch vectors in GPU.

Alternatively you can run old models with their GUI version, there's plenty that are free all over github (e.g: Releases · AaronFeng753/Waifu2x-Extension-GUI (https://github.com/AaronFeng753/Waifu2x-Extension-GUI/releases))
Hello! I used the chaiNNer app, and succeded with the HD spriting, here the difference. but im trying to get the sprite into the game and doesnt work, do you guys know why it is?
 

Attachments

Back
Top