zakhazard (@zakhazardVR) header
zakhazard (@zakhazardVR) profile picture

A doge but in vr. I do stuff maybe probably not. AI assisted

Joined July 2021
Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)Photo by zakhazard (@zakhazardVR)
Nuked LTCGI out of my projects we doing LVCGI now. Was originally thought to be 1/3~1/2 cheaper in practice though it was measured as being as low as 1/5 the cost.
1
5
150
And tonight fell for again award goes to.... Every developer who switched because of AI
1
1
222
Perhaps the greatest auto play of all time never deleting this app
This aged poorly
4
300
Mixed tracking still isn't as good as entire 2.0 setups but it definitely has improved. Having the headset communicate the wobble of rotors helped a lot. 2.0 trackers do struggle more than 3.0. There is also an option to prefer 2.0 switching only when it needs to.
New openhouse version released! Support for using 1.0 and 2.0 lighthouses together (read the README), even more improved tracking and room learning performance, support for Pimax headsets and numerous other fixes and improvements! Read the changelogs! github.com/YusufOzmen01/open…
2
19
954
Programmers actually love slave labor we call them Jr Devs
I mean yeah it's trained with slave labor, horrible for the environment, ruins the quality of life of marginalized communities. This is just another reason I hate it because it does all of that while being completely unnecessary.
2
137
Blender 2.79...
Now I know thats better create the scenarios in Blender than trying to do the same thing with cubes in Unity. #indiegame #gamedev #indiedev
1
6
249
It was legitimately grueling launching SteamVR as many times as did. And you just know it’ll instantly implode once users get their hands on it 🫩
upcoming experimental feature of openhouse: using 1.0 and 2.0 lighthouses AT THE SAME TIME! it is going to be VERY experimental and under a setting in SteamVR. hopefully will be available on the next v1.0.3 release! credits to @zakhazardVR for the feature and the videos!
1
2
20
2,252
Another 50% perf increase while further increasing quality. Now actually viable as a daily driver and not me just coping about it.
I actually had a good idea. Trained a NN(left) on a high sample reference allowing me to cut the sample count in half while recovering more occlusion in crevices. It is a bit noisier but can't beat a 50% perf increase.
6
490
Some bug fixes later here is an explanation on how it is possible and showing it does actually work. My base stations are also set up non optimal as a side note.
"I checked all the code" They are lying to not get crucified and thats ok. They left all their ai notes and one was about a mixed setup 1.0s and 2.0, but because they never pushed back against their ai they thought it wasn't possible.
4
7
107
5,614
Mind you the fact this works at all is crazy because the differences in timing and drift are measured in µs. There seems to be a driver limitation for headsets that force them into 2.0 mode. It can report back small timing changes so switches happen less so maybe better this way.
4
485
"I checked all the code" They are lying to not get crucified and thats ok. They left all their ai notes and one was about a mixed setup 1.0s and 2.0, but because they never pushed back against their ai they thought it wasn't possible.
4
3
2
151
20,891
He retweeted 1 vibe coded project that was useful and realized that looked like taking a stance
I've moved all work-related posting to @tuppervrc.
1
1
1
37
6,520
I actually had a good idea. Trained a NN(left) on a high sample reference allowing me to cut the sample count in half while recovering more occlusion in crevices. It is a bit noisier but can't beat a 50% perf increase.
I do enjoy just being able try a bunch of dumb ideas I feel like I have seen this before, but using good ol fashion machine learning to upscale a texture at runtime. Can recover sharper larger details and remove some compression but runs like ass*
1
1
3
803
Also this working makes me think DLSS 5.0 probably did start out as nueral screen space effects since NVIDIA has had their own replacements in the past (HBAO+). But they gotta sell their new gpus so make it something that run like garbage on old cards instead.
1
154
JPs gonna become anti ai real quick once they realize the cancer that is swarms that can just mass scrape ripper store and reupload their stuff
【⚠️ 類似・盗用疑惑商品に関するお知らせ】 こんにちは。Siru Houseです。 本日、情報提供をいただき、当ショップが制作・販売している**「Sheep House」の衣装と非常に酷似した商品**が、別のBOOTHショップにて販売されていることを確認いたしました。 該当商品は、衣装のデザインや構成など、多くの部分が当ショップの商品と酷似しており、現在、詳細について確認を進めております。 Siru Houseは、該当商品および販売者との間に、いかなるコラボレーションや販売許可の関係もないことをお知らせいたします。 現在、BOOTH運営側への報告および確認を依頼する予定です。 ご購入者の皆様におかれましては、誤認によるトラブルや被害を防ぐため、十分にご注意くださいますようお願いいたします。 今回、情報をご提供くださった方に心より感謝申し上げます。 今後もクリエイターとしての権利と、ご購入者の皆様からの信頼を守るため、誠実に対応してまいります。 Siru House 🔗 公式販売ショップ siruhouse.booth.pm/ 🔗 類似・盗用が疑われる商品 booth.pm/ja/items/9099859
2
7
1
90
8,229
zakhazard retweeted
日本人「欧米人は著作権守らないのに、なんでAI否定なんだ?」 欧米人「日本人は権利意識高いのに、なんでAI肯定なんだ?」
AIは嫌がっても利用規約は守らないんだね
17
81
7
778
62,305
How it feels putting an ai disclosure rather than just lying
2
203
I do enjoy just being able try a bunch of dumb ideas I feel like I have seen this before, but using good ol fashion machine learning to upscale a texture at runtime. Can recover sharper larger details and remove some compression but runs like ass*
1
1
2
570
While it does run terrible it realistically never actually does anything as the pink upscaled region needs you to shove your face into the texture basically.
1
101
One must imagine Richard Stallman happy
just open sourced ur shit pal😹😹😹😹 This is too easy
2
218