I only read about your Autostitcher after making my posts; I couldn't try it yet, but according to the comments I could read it looks like your tool is more useful than mine for this kind of maps. It's faster and it can work with higher quality images for that reason.
- How does it handle video at non-native resolution? The image matches might be on non-pixel boundaries.
- How does it decide what to do with artifacted video? Does it average them out?
I match the full image, not only the map boundaries... so video quality doesn't matter much, as I don't need pixel perfect matches (I do a image difference of the displaced images; then I compute the total error).
- What happens to sprites, animated backgrounds, etc?
Animated backgrounds are a problem. Sprites, not so much. It depends on the order I use to compose the final image: if I start from the first frame, you'll only see the player character in the last frame, since the only information that you can see for each frame that is not the last one is just a tiny border. For other characters, it depends. But their images are most of the times lost. It's hard to explain why, the point is that every pixel gets overwritten many times, and the last time it's written is the last time such pixel is seen in the video. Most of the times, an enemy character is not there then, there is at least some extra exposure for that pixel. Only if there is not the enemy character gets recorded in the image. If the order that's used for composing the image was reversed, though, the results would change.
It's also worth noting that as a rule, screenshot maps should always be produced in a lossless fashion. Having multiple lossy compression stages (video -> JPG -> JPG) must make things somewhat worse.
True. I thought this would be useful as a first draft for some games, maybe.
Like, say, you had a character that was nothing more than a hyphen (-), wouldn't the result turn out like this:
Not if the character was in the center of the screen, for example. It would be erased when all the frames were put together, as explained above.
Also, is this for Mac or something? I have no idea what GIT files are or PY or whatever.
Sorry, this is the Python source code only. No .exe provided still (but if you install python it should be easy to run it out of the box). I tested it on Linux but it should work in all platforms: Windows, Mac, Linux.