In case you're wondering, Create Border has been replaced with "Pad"

Really amazing tool, every updates gets a little better

I much appreciate you for doing this video. I use it all the time now and I probably would have a hard time without this technique.

THIS IS SO HELPFUL. THANK YOU VERY MUCH!

Great Video! It's so important too know, that most of the other Image Upscalers don't work properly (maybe such all of them). I'm glad I'm a perfectionist and registered that the images Topaz Gigapixel AI upscaled for me weren't seamless anymore... so I can now provide more value of my products with PNGs that are really seamless, even after upsacling, using Chainner!! Thank you a lot for this tutorial! I will download it right now and will finish the process you showed step by step. Seems to be a very powerful tool, even it's more complex. Thank you a lot! Gonna try this :)

Great vid

Hi, Thanks for taking the time to upload the tutorial. I am having trouble getting SD to work on my mac so I set about doing this manually and there is a key point that some people may miss (like myself). You need to increase the size of the document by 16px and fill the extra pixels WITH THE PATTERNED IMAGE, not transparancy or white space! Then upscale how ever you like and crop back the extra 64px from all sides. Workes perfectly! This is totally obvious when you think about it, but it took me two hours to work that out. Slam me if you will, but hopefully my time saves your time.

Hi, thanks for your tutorial, may i know what if i only want to upscale 2 times only(instead of 4 times)? I tried adjust the number in MATH box, but it didn't work, please help, thanks.

im sorry i might be stupid but where do i get the models from? my models folder does not have the same folders you do or the onse that i got are empty ...

I am on a 2016 macbook pro and the first pyTorch model I tried worked well but took 75 seconds per image. I used Chainner to convert the model to NCNN and this now runs the smae image in about 6 seconds.

batch upload maybe inconvenient for textures with different size.

What's the average computational time? I know this will depend on CPU and RAM of the machine, but would just like to set realistic expectations...

Such a great tool, but everytime I want to use it I can't, and have to wait for it to update python for an hour or three. By ther time it's done i dont' even need it anymore. Am I doing something wrong?

can we do bulk images here?

Can I use this same to enhance a portrait image of a person?

I keep getting "Failed to Fetch" error message on Mac. Any idea why?

Why do you upscale it two times ? Is there a reason for that ? Nevermind i found the answer of my question... There are some things you can do to "improve" your workflow: You should first delete the second upscaler. It's clearly not needed and performance hungry if you're upscaling pictures greater than 1024x1024. If you do so, you need to change the second and third "Math" to 2 instead of 4, if it's 2x model set the second to 1 and the third to 2. Also the Resize (Resolution) has purpose only if your Crop node is greater than your target resolution. Also i went into a loophole because of the upscaler's "Auto (estimation)" If you're stuck on it, change it to 256 or 512 size. It will prevent loophole (seen on the main.log). This should improve slighty the speed of upscaling for users with low-mid end computers. Very good workflow i never though it would be possible to actually do seamless upscales with perfect results.

Hi! Thank you for the tutorial! I try use it on Mac. In the Dependency Manager I see that the AUTOMATIC1111/stable-diffusipon-webui Not Connected. How can I connect it? :)