Convert image to PETSCII

[Esshahn]

(it had to come.... :D )

• load in PNG file
• could be either 320x200 or internally resize to that format
• converts every 8x8 pixel area into the closest character pattern match
• same for colors

If I understood correctly, your rotation concept already uses this pattern matching, right? I'm not sure how to identify which background color to use though.

[nurpax]

I have many needs for this feature so rather than write this into some private python script for my own use, figured might as well stick this into Petmate. So moving to 0.5 milestone.

[Esshahn]

yay \o/

[nurpax]

Mostly done but this still needs:

• Docs (explain caveats as this conversion doesn't always work)
• Add "select which background color conversion" to allow users to pick which conversion result he/she wants. The thing is that some PETSCII PNGs are ambiguous in the sense that the same PNG can be a valid PETSCII image with multiple different background colors. My code returns all of these but I now just pick the first match. I'd like there to be a selector for this so that a user can select which background color to use.

[Esshahn]

So I've tried it with the beta 0.5 without any luck. For my better understanding: could it be that the importer only works on PNGs that are already petscii? Because whatever 320x200 PNG I threw at it, I always got the red colored error message.

This is what I did expect:

• select any resolution (but 320x200 is okay for now)
• select JPG or PNG (but PNG is okay for now)
• ability to move around the image on the canvas (which would give very different results)
• convert into petscii

Like this

[digithree]

Is there any idea for how to proceed with this? I have started looking at a possible implementation of a more tolerant import tool but I don't want to double up on work.

[nurpax]

Will probably still take a while for this to materialize in Petmate. Been a bit too busy at work and haven’t worked on my hobby projects much.

I have an idea for a UI that’d allow setting the right crop and scale for the input to guide the conversion process. I don’t have a good robust algorithm in mind for a fully automated converter. I think it’s possible but I’d prefer a human guides converter over a fully automated thing that sometimes fails.

If you have some ideas for doing this well, please share. :)

[nurpax]

Oh btw my comment was re: converting a screenshot of PETSCII to an exact PETSCII screencodes and colors. The algo that @Esshahn talks about is different. I do have a reasonably good converter for this written in Python/numpy and it’s not difficult in JS either. But needs the UI which takes more work. Not rocket science but anything I do that involves any CSS layout takes me a lot of time.

[earthbound19]

@nurpax this isn't directly relevant to this issue, but I'm curious: what converter do you have written in Python, and is the code online, and if so, where?

[nurpax]

@earthbound19 Hi! I never released it publicly. I don't know if it's really all that good tho. :)

[earthbound19]

Ok.

I'm now imagining a more general-purpose converter (which I might like to develop) that can reduce any image to any visually matching arbitrary set of glyphs (e.g. from any font), and any set of colors. That could cover commodore and a whole host of other older and modern computers.