Hacker News Reader

Comment by maggit | Hacker News Reader

I was excited about this when it was all the rage. It speaks to me as a programmer. Some years later, I stumbled upon this excellent article by an artist considering the problem from the art perspective: http://www.dinofarmgames.com/a-pixel-artist-renounces-pixel-...

It points to the problem with these upscaled versions: They aren't any good. In the end, the artist demonstrates how to get good results, draw it again: http://www.dinofarmgames.com/wp-content/uploads/2015/04/yosh...

Of course, the algorithmic approach lets you upscale any number of preexisting games. However, I have concluded that I like the original art better.

View on HN

That article is fantastic, so right!

There's a ton of graphics research that has the same basic issue, I feel. The premise behind a lot of it seems to be how to do impressive art without an artist, how to automate our way to compelling imagery.

I am guilty of this too, as an occasional graphics researcher and full time tool maker, I'm implicating myself here too. I am with you, the algorithms speak to me as a programmer, and they are really fun to think about and code.

But I suspect we're missing out on some opportunities to do amazing things with an artist in the loop, that we're not yet achieving what we could if we built tools that try to enhance creative people rather than replace them.

Adobe is pretty good about taking Siggraph papers and implementing them with more artist friendly controls than the paper, but I can't help but wonder what we might have if all the Siggraph papers on average aimed to be more controllable and less automatic in the first place.

In their comparison http://johanneskopf.de/publications/pixelart/supplementary/m... the Vector Magic results appeal the most to me precisely for this reason. For most images they look more like an artists rendition than an attempt to trace the image. Which is good.

when it works it's beautiful but worse cases like mario face are ruined. hq4x isn't as pretty but seems more consistent over the whole gallery of examples http://johanneskopf.de/publications/pixelart/supplementary/m...

I had to look but there is a NES emulator that does sprite replacement which is interesting:

https://github.com/mkwong98/HDNes https://www.youtube.com/watch?v=N57ApspYHVE

>Of course, the algorithmic approach lets you upscale any number of preexisting games.

This is the point of these algorithms, for sure. Hand-crafted graphics will always beat algorithmic upscaling of lower quality hand crafted graphics, given the same amount of time to do each, but that isn't possible for most examples of games with pixel art.

Reply on news.ycombinator.com