oopsie I forgot for a bit

I've been getting back to work on Macrovision content! This includes new characters (some from the backlog, some freshly submitted) and a whole new category: real buildings.

Now, I obviously do already have some real buildings, mostly under the Landmarks category. This time, however, I'm working with models from the Open Street Map project. This lets me import an entire city's worth of buildings.

The path from Blender to Macrovision is a bit convoluted, though:

1. I run a script to isolate the building I'm interested in and take five pictures of it. Three pictures are taken from the side: north, northwest, and west. Two pictures are taken from the top: north and west. The world is unlit, so the model itself is dark, but the Freestyle renderer is used to draw lines on the edges.
2. I batch-image-trace these images twice. The first batch traces regardless of color; the second batch ignores the white lines.
3. I batch-convert the resulting Illustrator files to finalize the trace and to, for the second batch, adjust the color to a dark gray. This results in .svg files.
4. I run a script to manually combine the .svg files together. This is cursed.
5. I batch-convert the resulting combined .svg files to auto-trim their bounds

This would be much simpler if Illustrator's action recording system could remember my image tracing settings...but alas, it cannot.

So, yeah, that's cursed. The process has several manual steps, and takes a while, but it's mostly hands-off. So far, I've added ten buildings from Chicago and eighteen buildings from NYC. I'm currently working on a big pile from Houston.

These buildings should give a much wider range of things to compare against for modestly sized macros. You'll find them under the real-buildings category.

Beyond new content, I'm looking into a few lingering bugs. The view should no longer jump a bit when panning after zooming with the y-axis unlocked. I'm thinking of making zooming happen relative to the mouse cursor, rather than the center of the screen, but that'll take a bit of thinking (and MATH).

Rotation is also something I'd really like to add. Actually rotating the images is trivial; the problem is that my click detection gets royally fucked up, and I've had a surprising amount of trouble fixing it.

Clicks are detected by drawing the entity to an invisible canvas and sampling it at the point you clicked. If the pixel is opaque, it's a hit. This works excellently, but getting the rotated image placed correctly has proved elusive. It's always rotating around the wrong point, or winding up flung off the canvas. I'm sure I'll get it sometime soon! The same problem affects flipped images, incidentally.

---

I've been playing a lot of VRChat. Whee. Been playing around with making worlds. Nothing much to report yet, though.

---

I mentioned a few weeks ago that I was looking into making VR content. I've made some great headway in that department -- including a few trial runs of ambisonic audio. Ambisonic audio captures an entire sound field; it's a bit like how a light field camera can be refocused in post, and lets you listen at any angle.

Unfortunately, I have yet to find a player that correctly handles both ambisonic audio and that can play 360 stereoscopic video in VR. All of the VR players I've tried don't handle the sound properly! VLC can play 360 video and handles the ambisonic audio, but it can't display in VR (yet).

The stopgap solution is, apparently, Unity. That lets me get everything working at once. I might have something finished pretty soon (along with a regular 360 version for those without VR).

technology sucks lol