Have you ever seen a post where it's a photo of someones screen instead of a screenshot? This lora is the solution to that. It re-frames the image so and removes scan lines giving a screenshot like output.

The output will almost definitely need to be upscaled as the original level of detail cannot possible be achieved. Also I only trained on one type of screen so I'm curious is it will work with others. Also it should work on phone screens but i am not sure how well considering i only used a single photo of one in the dataset.

I kind of trained this just to see if it was possible so I wouldn't expect great results but with a good upscale I'm sure you all could get some decent stuff out of it.

Use the prompt: convert to screenshot