>>1020530
>is this actually more performant than the existing hair nodes?
Honestly? I don't know. I haven't crafted a full head of hair using the existing nodes in a long time. I've started a number of times. But get bogged down with all that's required to make those nodes work. Dunno why. They just don't suit me.

For me, this is about ease of use. Doing the base mesh like this settles many issues. Interpolating between curves is perfect. Creating parts in the hair is easy. There are fewer points to manipulates in order to shape the hair, so you're not grooming a thousand parent curves. If nothing else, I'll pursue this further just for the ease of it.

Since I'm on my shitty PC(my good one broke), I can't really tell how fast it is on a real PC.I have noticed that it's *relatively* faster to create points, and then convert points into curves. Rather than create curves first and instance them. So I believe that setting it up the way I did is faster than the typical instancing method most tutorials teach.

>production grooms are in the 100k curve range
With the mesh disabled, Blender handles curves a lot better. Creating 100k curves takes about 1,300ms up front. However, the viewport has no lag when navigated(pic related), and there is only a slight amount of lag when I toggle options. Running the timeline is 1 frame per second however. So I guess that 100k curves would be light work on a real PC. It's the mesh that adds the most amount of time.

>is this based on on the yuskel papers?
Yeah. In the most loosest way possible. I just skimmed and got the gist. Creating the base mesh looked doable.

>i thought the performance gains for those ones came from building curves directly on the gpu
That's right. I don't know how to do that part. I experimented with SDFs a while back and got interesting results. But it didn't scale as well as I wished.
>>1004257
>>1005159