chaiNNer VS upscayl

What are you trying to upscale? There's a whole lot of different upscalers out there, each best used for a specific type of image. If i do a cartoon character i want to blow up, i use RealESRGAN 4xAnime (6B) and then throw the upscaled image in Gigapixel AI and blow it up 6x. So in the end i get a super sharp 12k+ image out of it. You can download an upscale model and use it in Chainner on your local machine or throw it in the ESRGAN folder in your AUTOMATIC111 SD fork.

using ChaiNNer - https://github.com/JoeyBallentine/chaiNNer

I thought about doing something similar, but instead of creating my own project, I would have made a module for this one: https://github.com/joeyballentine/chaiNNer

For these, I used the software gigapixel AI. A free alternative is the process described here https://youtu.be/g_FlJ9yIG9U using this software. https://github.com/JoeyBallentine/chaiNNer

Alternatively, you could keep using your current model and correct the color shift. ChaiNNer has 2 nodes that can do this called "Average Color Fix" and "Color Transfer". Both nodes take your upscaled image and the original image (reference) and try to make the upscaled image look like the original image by shifting colors.

Here is the link https://github.com/joeyballentine/chainner and you download the models for ESRGAN from here https://upscale.wiki/wiki/Model_Database

Use ChaiNNer instead. Cupscale is outdated and hasn't been updated for a year.

I use this, https://github.com/joeyballentine/chaiNNer

The video embedded on the README shows a pretty decent example of the tool in action with a before and after. The code I could find uses PyTorch neural networks https://github.com/TGS963/upscayl/blob/v1.1.1/scripts/test.py so maybe so on the use of ML, but I haven't seen yet what's in the source code gzip. In any event this is nice work, kudos!

Cool. Where's the source for https://github.com/TGS963/upscayl/blob/main/resources/linux/bin/upscayl ?