Ngonna make a scrollbar for it
Nwhere the stuff either emerges or gets sucked into the pot
Noh whaaat that'd be cool to graph
LYeah!!
(Not cool at all when you have to try different nixpkgs commits until one unbreaks your system to try to find exactly the culprit tho…)
(Not cool at all when you have to try different nixpkgs commits until one unbreaks your system to try to find exactly the culprit tho…)
NNreally good article that i should remember to do more when i bash on visionOS and such
Nbecause it's got problems but they're not quite as bad as i make it seem in terms of "the user cannot use this thing well at all" like yea it needs context switches but that's not all doom, it's just a bit more friction to overcome (still makes the device useless in practice tho lol)
Nthe final conclusions i draw i stand by even thinking through it deeper
BInferring from LOS hand tracking is a pain, not least because it's LOS
NLOS?
Bline of sight
Nahh yeah
NvisionOS' real problems with gazepinch are midas touch and lack of contextually rich input
BI am partial to gaze cursor as a quick and dirty solution tho, it's darn fast
however
"you leave me no choice! I cast slightly-misaligned eye tracking!"
however
"you leave me no choice! I cast slightly-misaligned eye tracking!"
Nit's a generalized solution across all of 3D space that doesn't generalize well
Byeah :/
Nwhen you have heuristic-based input the best way to improve its accuracy tends to be to contextualize it
Ngot hand tracking where multiple controls could be tapped? change the controls to require different interaction methods and now it'll be harder to mix them up
Nthat kind of thing
BVery much yes
Since it wouldn't be too uncommon to want to doodle in fron of a window, for instance
Since it wouldn't be too uncommon to want to doodle in fron of a window, for instance
Nthat's what comet does well
Nbeing able to just curl your fingers near the comet pen to make a pose like you're holding a pen to hold the pen
Nok bad wording but like
Nit's uncommon and extremely fast
Ni need better signifiers for it tho
Nright now stardust has a serious lack of signifiers for stuff
Nmostly because no custom shaders and limited geometry support (no blendshapes :(()
Ni'm not a graphics programmer haha
BI haven't fooled around with comet at all since it was a little broke when I did my big fiddle bonanza
BIs there any reason why custom shaders wouldn't work, or is it just low-priority?
Nbecause idk how to make shader code work properly lol
Noh and also i'd need to jam the sk shader compiler into the binary and compile stuff at runtime
Nor jam in a GLSL shader into the raw pointers
Nwhich.. idk how to do
Nit's possible!! but i have tried graphics programming and it'd simply take me too long to do it in practice
Ni have other stuff i know i can do like UX interaction that's much more important than signifiers atm
Ni can explain how a thing works because i only have power users and devs rn who are used to overcoming friction
Nas long as i get the good signifiers and design in before i expect others to use it, it should be fine
BAh, no worries
I went down the shader rabbit hole when I was working on the Weave-chan avi earlier this year, though that was in Unity
As for how to handle the interface between shader script and server, idk
I went down the shader rabbit hole when I was working on the Weave-chan avi earlier this year, though that was in Unity
As for how to handle the interface between shader script and server, idk
Nwell, i'd be using something akin to MaterialX
Na high level shader file format that can map to whatever memory addresses and shaders that know the lighting from the underlying rendering pipeline or whatever
Ni don't wanna force WGSL or GLSL or such because that'd mean you must use 1 particular rendering pipeline for stardust
Ncan't reimplement servers or such
BGraphical things and whatnot done activate me neurons
BOh man I forgot about that problem
ahem Unity
ahem Unity
Nunity actually does this too
Nthat's why ShaderLab exists
Nto make shaders work across rendering pipelines
Nit's similar syntactically to HLSL but it's not HLSL