checking out Easy diffusion

Tried out Easy Diffusion, https://github.com/easydiffusion/easydiffusion

checking to see if it is faster than Stable Diffusion webui - my previous post on that is here.

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-NVidia-GPUs#linux

Some useful links - basics of inpainting - https://github.com/easydiffusion/easydiffusion/wiki/Inpainting

https://allprompts.com/stable-diffusion-prompts-for-realistic-photos/

So we should add words like national geographic, landscape, photo, photography

Easy Diffusion - better interface, and a bit faster - 45 sec instead of 60 sec to generate a 512x512 image, and just 7 sec to upscale the generated image. But there doesn't seem to be a way to upscale our own images, at least not seen in the UI. Trying the same thing with CPU on laptop, the 45 second generation on the 1060 graphics card took nearly 12 minutes with CPU!

Some example prompts and resulting images:

1. moonlit landscape - 

2. photo of night sky with full moon tokyo skyline in silhoutte nikon national geographic

(to insert the generated image here.)

3. moonlit landscape, nikon, national geographic, full moon, trees (wish upscaling done 512 to 1024) -