Hacker News Reader

I've dreamed about doing an automated "ASCII"-art converter that utilizes all of Unicode. You could accomplish some pretty crazy line art if you put your mind to it, but the analysis would be super challenging.

Also you'd either need a monospace Unicode font (which I don't imagine has anywhere near 100% glyph support) or incorporate spacing and kerning rules...

View on HN

I would think about feeding rendered unicode characters to train a visual algorithm about which character may match, then "on the go" do a bitmap comparison to select the best one for the given case.

The first part could be something like shapecatcher.com where you select the top N plausible candidates, then on the final part you select the best one for the case at hand. (well at least that's how I would try to implement that first)

Anyway yes, it would be great!!

Reply on news.ycombinator.com