Azad Balabanian (@Azadux) header
Azad Balabanian (@Azadux) profile picture

3D scanning/VR—formerly @nianticspatial, @puzzlingplaces, @uploadvr. Perpetuel typos

San Francisco, CA
Joined November 2010
Photo by Azad Balabanian (@Azadux)Photo by Azad Balabanian (@Azadux)Photo by Azad Balabanian (@Azadux)Photo by Azad Balabanian (@Azadux)Photo by Azad Balabanian (@Azadux)Photo by Azad Balabanian (@Azadux)Photo by Azad Balabanian (@Azadux)Photo by Azad Balabanian (@Azadux)Photo by Azad Balabanian (@Azadux)
Today, I did some aerial photogrammetry from a plane! Due to the cabin being tight and my cam’s viewfinder disp. not rotating, i had to get clever. Using the HDMI input on my @FatsharkRC FPV goggles, I had a clean feed of my camera, which worked super well!
17
29
13
363
Azad Balabanian retweeted
Fellow Gaussian Splat fans, Frankensplat is out! * A new 3DGS training tool & engine, written entirely in JS & WebGPU * Up to 2x faster than LFS at comparable quality * Fully browser-based * No cloud, no downloads, no installs, no builds, no EXE files Links in reply. Enjoy!
My fellow Gaussian Splat enjoyers, I created a monster. I'm about to set it free this week. Be prepared.
6
26
3
441
16,184
I wanna get better at writing... both for video scripts and blog post Oftentimes, I have a core idea (like a tweet) but I feel rusty when it comes to structuring the full thing I'm starting to use @mattpocockuk's grill-me like skills (writing-fragments and writing-beats) to flesh the ideas out and give them structure-- feels like im onto something! any other suggestions?
1
6
565
Azad Balabanian retweeted
Introducing Frame.
24
22
19
585
58,334
Azad Balabanian retweeted
👉It's been a while since we had Insta360. My little shadows, they found something.
2
6
1
74
5,340
ive always loved multiview/lightfield displays for IRL content here's the opera house in yerevan, on a gen1 looking glass display by @nikkiccccc (i think?) from 2019
1
1
1
19
1,124
Azad Balabanian retweeted
I trained a text-to-speech model that has never heard a real human voice. 4.5M parameters. 18 MB. Runs on a CPU. Meet tiny-voca
291
541
116
8,339
505,641
Hey guys check out this incredible video Oh wait it’s derivative slop because it lifts the iconic intro music from House nvm it’s trash
unavailable
1
1
939
ive been a codex guy for the last 11 months. this month, i got claude for opus 5.5 and fable and am very much enjoying it I do miss the familiarity i had with codex as a harness though. there's a learning curve to the claude app i might try out t3 code soon though so many it wont matter
28
39
6,255
Azad Balabanian retweeted
SplatBox v1.3 🚀 Grab it here: simulacrum.dk/Shop#unity-gsp… New dark horizontal UI + Splat Decimator. Paged LoD uses spatial streaming, Burst traversal, GPU depth sorting and refinement by camera distance. Persistent edits + Relight support. A 7.3M splat scene loads on Quest 3! #SplatBox #Unity3D #3DGS #LoD
2
6
41
2,415
Azad Balabanian retweeted
Fable 5.2 wrote the transformer. Opus 5.5 generated the training data. The result: a 367K parameter language model that fits on a floppy disk and runs on a Pentium. AI is now building AI for hardware from 1995.
Readers added context they thought people might want to know
indian engagement account making it seem like this is his when he just took someone else’s post from reddit without credit (for x payout money) reddit.com/r/ClaudeAI/s/m…
177
386
111
6,522
492,697
Azad Balabanian retweeted
TAKSIM SPIRAL (2013) Taksim Square is one of Istanbul’s most significant public spaces and has been the epicenter of many important social and political events, including the Gezi Park protests of 2013. On the morning of November 13, 2013, I drew a spiral with a piece of chalk on the ground in Taksim Square and performed as a working-class man. Listening to a metronome through headphones, I walked along the spiral path for approximately 10 minutes. By compressing the entire duration of the walk, the action was transformed into a rhythmic moving pattern lasting less than a second. The work is part of the Clones Project, a series of video performances in public spaces in which repeated actions are condensed into looping choreographies. At the time, these pieces were published as GIFs on my blog. I later exhibited them internationally as editioned high-resolution video works.
4
20
1
252
52,782
Marking this as a historical moment where a non-meme ai gen visual has broken into the zeitgeist Also, it’s not just ai gen. He’s doing incredible things with the sound design, the poem edit, the world building, the video editing, all of it. Who knows if there’s a movie behind it. Doesn’t matter really
NUREMBERG - Official Trailer
21
4
1
178
14,090
I was spot on
My personal thesis is that As images and videos becomes cheaper to “generate”, the value of high-effort art will increase It’s analogous to what’s happening to movie marketing today with the “no CGI was used, it’s all real” When CGI is so cheap to use everywhere, people start to value the high effort productions more, hence why Top Gun, Mission Impossible, and F1 leaned so hard on the “this was filmed forreal”
2
9
852
Azad Balabanian retweeted
We just released Solv3R on 🤗 HuggingFace Space! 🔗 huggingface.co/spaces/jgkwak… Try our interactive demo with your own images and see how Solv3R reconstructs your scene in 3D! 🚀
📢📢📢 Introducing 𝐒𝐨𝐥𝐯𝟑𝐑 Can geometric foundation models move beyond fixed-set prediction and keep solving like classical SfM? We turn a looped Transformer into a reusable 𝐥𝐞𝐚𝐫𝐧𝐞𝐝 𝐠𝐞𝐨𝐦𝐞𝐭𝐫𝐢𝐜 𝐬𝐨𝐥𝐯𝐞𝐫 🌐 Project: ubc-vision.github.io/solv3r 🧵👇
1
7
3
44
5,252
Azad Balabanian retweeted
You really, really need to be more ambitious Things are gonna get really strange over the next year, and if you're not using these new models to do bigger greater things your going to have a very very bad time Working on my Fable 5.2 vid rn, the runs keep going longer and better than expected and it's making me believe this even more. We have barely scratched the surface of what these things are capable of...
24
25
5
583
21,509
Azad Balabanian retweeted
A Gaussian splat of a blue morpho butterfly. It has a remarkable blue shimmer and it's actually huge, around 12 cm. I had to adjust my setup to capture it. #3dgs
14
43
5
611
13,081
Azad Balabanian retweeted
lasertag can be a horror game (ft @kevando_)
5
6
1
80
4,703
Spirula Studio is just.... so fking good as a Gaussian splat training software, esp for 360 cameras it does everything. image alignment, mask generation, normal/depth map estimation, auto scaling, and splat training, all in one exe. Runs on Windows, Mac, and linux too, and doesnt require CUDA I'm working through on finding the optimal settings. Once I do, expect a guide + tutorial
8
11
144
4,863
Azad Balabanian retweeted
OpenVRLink got approved for Testflight 🎉 OpenVRLink is a fully open-source implementation of the Steam VR Link client protocol, running on visionOS, with working dynamic foveation and (limited) face tracking
11
23
2
311
10,746
After 30+ years in 3D, I just released my first ever app: RAYS, a portable path tracer for iPhone Duo, iPhone and iPad. 🎉 It's the first in a suite of creative apps I'm building. stevetalkowski.github.io/ray… #3D #pathtracing
13
15
3
98
5,103