upscaling and resizing with chainner

Chainner - A node-based image processing GUI - seems to have a relatively easy-to-use installer - deb files for Linux, Windows installer, etc. And it has a dependency manager which can install the required dependencies. Tried it out, but unfortunately, it could not use the external stable diffusion api - maybe because I didn't run the web ui with API enabled? or is it that the API has changed and hence it gives errors like "invalid HTTP request"? Tried running chainner after running easydiffusion - could not connect.

api wiki says latest info is at :7860/docs

http://127.0.0.1:7860/docs

Video - how to upscale images with chainner - 

Models from https://openmodeldb.info/

(old site was https://upscale.wiki/wiki/Main_Page#The_Model_Database )

Chose https://openmodeldb.info/models/4x-RealWebPhoto-v4-dat2

Took 10 minutes. Seems to be using cpu and not gpu

8x upscale with onnx took 24 minutes. Looks like water colour and not pointillist like the previous upscale with Stable Diffusion. 

8x upscale with onnx using the model 4x-RealWebPhoto-v4-dat2 compared with the original image before scaling down and upscaling. Click on the image to see it larger.

Probably there may be some models which would do fast upscaling with realistic results (unlike the compact model below which seems to be optimised for anime). 

Recommended models at

https://phhofm.github.io/upscale/favorites.html

Compact model - https://openmodeldb.info/models/2x-HFA2kCompact

For video, the youtube video above used video frame iterator, and the compact model. Now the interface has video input and video output nodes, which will automatically create video from the frames. Using the  compact model, the video upscaling and processing was completed in just a few seconds, for 250 frames Baba. Looks a bit like anime / cartoon / water colour. If the numbers under the chainner nodes are the times taken, then 0.04 sec per frame, 9.8 sec for encoding to vid.

Edit: There are lots of youtube tutorials for video upscaling, including ones using colab - https://www.bing.com/videos/search?q=stable+diffusion+video+upscaling