{wonder, innovation, elegance} ∈ I turn emerging technologies into award winning apps. Ex-Hackathon pro.
#3D
#AR
#AI
since forever. Mad science and artistry ❤️
Because SOTA
#AI
drops daily: the future of computing is all about
#HCI
- the most intuitive UI UX possible between human and computer.
And also little point in incrementals on existing models when behemoths like
#openAI
will just periodically dump mega improvements
Missed an updated Shap-E from
#OpenAI
#devday
releases - so here's an
#AI3D
app demo using just a mix of
#opensourceAI
models
Napkinmatic3D "Toy Ripper mode" lets you rip any product image into 3D printable parts.
What if you can create a 3D Vision Pro volume app by simply chatting with it?
VolumeMatic Pro lets anyone create 3D interactive games for visionOS.
This is not SORA and not discord. It's a new app I'm building - evolution of AI3D projects I've demo'd at
#SIGGRAPH
and beyond
AI3D Camera Obscura lets you turn a photo into its own world by AI converting detected objects into 3D models - on Apple Vision Pro - by being skeuomorphicly curious on VisionOS!
It's my remote proj for
#VisionDevCamp
Thanks to
@tipatat
for the fun
@visiondevcamp
photo
How to use one
#AI3D
service in
#Napkinmatic3D
“without an artist”:
take a photo of anyone’s face or just find a portrait photo online… and you can still use concept art idea buttons that you scrawled on paper to AI re-design your 3D XR AR avatar
#aitinkerers
#SF
remote
Experimental XR Theatre: A live virtual actor (as a 3d clay avatar) is created from a Vision Pro persona - in a spatial computing auditorium, inspired by the Globe Theatre.
#HappyPiDay
- other than the input photo, all content is generated by AI - it's an AI3D Theatre in the
Wore the AI3D hacker hat today!
My experimental Vision Pro “spatial computing native 3D creator app” that turns rough 3D blocks to a fleshed out AI3D model (Minecraft-ish voxels render as LCM input for 5s image to 3D) won first place at
@aiengfoundation
realtime multimodal and
Filmed live from my waterfront coworking space in the Presidio:
This visionOS AI creator app is an example of a tech demo vid that lacks structure and editing, aka, the raw capture. It's too long, I ramble, etc.
I have so many raw demos like this that lag behind time available
Debugging my ear on my Vision Pro with a Bluetooth endoscope that just streams to visionOS (Amazon link below)… raw buggy ear gunk beware - earQuery source code included 🤣
tldr; this junk is from 48 hours of nonstop AirPod pro wear - air out your itchy ears!
"X Like Me" - an experimental AI art installation that lets you "Walk into a Stonehenge of different diverse versions of you."
Starting with your Vision Pro persona - transcend race, gender, and (even) media. Connect with different forms of you to experience being "X Like Me".
Do things like this happen outside of SF?
Monday - toured ridic condos, negotiated 20% off lease with 1 year prepay at $100k
Tuesday - last min demo / 900+ attendees
Wednesday - discover a major $$$$$$$ hackathon is happening at W Hotel
Thursday - win $100k at said
AI3D Desktop got accepted to CVPR Demo Track!
(Even though I have been building apps using computer vision since 2010, this will be my first actual attendance at CVPR!)
Seattle for summer solstice!
Too hungry and foolish and crazy* for
#TEDAI
official stage rn - raw hacks from the future live onstage at 109th SF Hardware Meetup later today (Tues) 10.17 @ 630p Studio45
i'll talk about current XR AI methods for (non-BCI) computing at the speed of thought - and demo my new
AI3D Desktop has some sweet portable office toys.
Sneak:
@BlockadeLabs
Gumball Worlds Machine lets you coin op, pinch-the-handle, and output a grabbable expanding random "gumball world" skysphere.
It's the sweetest way to create new worlds on Vision Pro
"The Billion Dollar Photoshop Comp"
#timeless
#priceless
lets us teleport extinct animals and lost treasures ... in
#XR
using
#AI3D
... on my iPhone - and turns
@Photoshop
comps into windows into new universes
Live demo. Added
@BlockadeLabs
to Napkinmatic3D "Toy Ripper" - also
@Carnage4Life
In the flip case, even when you are the largest contributor and spent most of your life developing the skills to build it on Uberman schedule... you're told to try to tone down what you did and say it's a team effort.
whenever i go to founder dinners and there are like 8 people in the table and everyone introduces themselves as, "Hi, I am _ and raised XY million" - i feel like i need to grow up, "Hi, I'm Yosun. I'm grateful to be here, but I've never raised VC because the terms have just never
Biggest challenge yet. No idea how I'll do this - at least not the traditional way since right now, what really excites me is building a new AR hack...
It costs the price of a
#Quest2
to buy a $299 Developer Strap for USB C data I/O to the Apple
#VisionPro
.
Can't wait for the gucci decked $2999 version of this dev cable.
(I feel super weird receiving a headset that doesn't let me cable connect to it out of the box. Is being
Using spatial computing as a means for humans to easily communicate with AI on Vision Pro
Blocky Minecraft AI3D Mesh concept to image to Mesh aka 3D + text → 2D → 3D
Glad to win
@aiengfoundation
hackathon - gotta work on being better at telling part of pitching the future!
Continuing to explore skeuomorphic spatial computing design for VisionOS with what started as a random hack while 3X Audibles through Isaacson’s da Vinci biography
AI3D Camera Obscura project WIP for
#VisionDevCamp
What do you think this app should do on Vision Pro?
One of those weeks when I can only say… Mercury is definitely in retrograde.
Sunday - tragic breakup
Monday - AI3D got rejected from SouthPark Commons
Tuesday - AIGrant rejection
Wednesday - OSV rejection
Thursday - a startup I’ve advised mostly on equity despite
My AIGrant founder video filmed as my Vision Pro persona (VisionOS 1.0) was so bad they rejected me without even checking out my demo :-(
In the age of easily doctored videos where you can be any race and gender - next time, I’d apply as my inner white dude
tldr YC founder
So much fun demoing
#Napkinmatic
spatial computing ⇔ AI controlnet-scribble-sd1.5 (replace endpoint with your own!) at
#WoodstockAI
- photos from moments before the event - with
@corgi_cam
… and my AI team
Why is there consistently no innovation in game ads?
#GDC2024
This is what in-game ads could have been 10 years ago, but sadly AI video beat it:
1) imagine if Unity Ads showed up as Texture2Ds instead of bannerads and popups
2) get eyeball time from camera culling when
Wow! Worth watching this. China's largest video platform
#Tencentvideo
(97M paying China subscribers) will begin inserting extra ads into movies/series that didn't exist in the original.
#computervision
New intro: Hi, I'm Yosun Chang, a failed entrepreneur who hasn't been able to figure out how to monetize from mad science and artistry (produced through self-enslavement, of course!)
✅ Win 3 million dollar hackathons
✅ Win 2 TechCrunch Disrupt Grand Prizes
✅ Win Best of SIGGRAPH
One day, when I grow up…
MacArthur Fellowship
XPRIZE
Nobel Prize
AI3D Desktop got accepted to CVPR Demo Track!
(Even though I have been building apps using computer vision since 2010, this will be my first actual attendance at CVPR!)
Seattle for summer solstice!
Activating ET's Magic Touch on
#Hololens2
- a quick
#MRTK
hack I made at
#MRDevDays
... or quantum leap 2019 in doing a finger to finger on an
#AR
hmd!
Longer Youtube:
Hey, it’s me and my latest magical app onstage vs billion dollar research company apps again…
#Napkinmatic
will be live demoing as one of the best real-time entries at
#SIGGRAPH
#RealTimeLive
(Also I won’t have to bum your badge since this contrib comes with 2 free full
@EMostaque
Or... what if you could have your own AI 3D AR character take you to posters of interest to you and explain it, translated "in language of your own latent space", even when the topic is left field and the presenter isn't there?
@SMBDYai
crashing
#NeurIPS2023
#AI3D
This is going to be so fun to win.
tldr; Been waiting for this forever. 20 years ago did some NSF funded quantum computing research, invented a bunch of qApps but decided not to publish because of the depression of none of it being real or useful within lifetime. Winning 5M
my rapid prototyping language for today's OpenAI DevDay's crazy hack - aiQuery - is somewhat inspired by
@jeresig
's jQuery
// Chain GPT4o to do text → image → 3D
$(ai).model="gpt-4o";
$(ai).text2image("INSERT_PROMPT").image2mesh();
and now to pause hacking to make that
Missed an updated Shap-E from
#OpenAI
#devday
releases - so here's an
#AI3D
app demo using just a mix of
#opensourceAI
models
Napkinmatic3D "Toy Ripper mode" lets you rip any product image into 3D printable parts.
What if you let
#OpenAI
#GPT4o
design its own face, so that you can teleport your AI into the real world as an embodied being?
#AI3D
headsim frees your AI from the chatbox, so that you can experience AGI in real life - the way she/he/it intends to look and sound like.
too quick at building stuff. too much free money out there. contest winning potential in January from having spent just weekend time: $2500 + $27,500 + $20,000 + $15,000 = aka how to
#bootstrap
your
#startup
from
#hackathons
@michaelryap
the fun part about this video is that these days, it's not hard to {scene-identification}-{object-segment} a video then use
#LeapMotion
input for {hand-cursor} {localized video-shader-streaks}
Personal gauntlet to win another one at SIGGRAPH in August this year with Napkinmatic 3D: sketch+ideas on paper to (textured) 3D model.
Names confuse. It’s funny that
@DrawmaticAR
was a domain registered for a 2019 model that couldn’t quite do what Napkinmatic 2D/ControlNet
Prompt to low poly 3D mesh in 7s… in an app you can grab on your phone.
Went the experimental wanderlust monetization route with
#PolyMagical
, partly due to a Square hackathon. Mixing 3D Gen
#AI
with
#AR
in a
@buymeacoffee
style platform but with travel instead.
@maxweisel
Should have been Magic Leap “Third-Party” Porting Grant instead of Indie Creator - latter has a totally different expectation. Being someone who has been building AR apps for the past decade, instead of pitching new concepts, would have just stuck with HoloLens apps I built.
My experimental Vision Pro app that uses the latest in computer vision and AI to let anyone to explore being a different race and gender on a spatial computing easel... got rejected from CVPR Art Gallery :-(
"X Like Me" - an experimental AI art installation that lets you "Walk into a Stonehenge of different diverse versions of you."
Starting with your Vision Pro persona - transcend race, gender, and (even) media. Connect with different forms of you to experience being "X Like Me".
Very excited that
@mrcatid
will be building the production version of Drone Vision AVP with 60 fps AI upsampling livestream DIRECTLY from your drone to Apple Vision Pro.
Check out and
Drone Vision gives you wings.
3/ Drone Vision
Whisper + function calling to interpret voice commands and navigate drones
(RIP the drone which broke during the demo 🪦)
@Yosun
@sagarsaija
🥉3rd place— $500 cash
Tried building a game changing ML-powered accessibility app with tinyYolo binding directional audio at the ATT Hackathon... but ended up winning Best Entertainment app (non game).
#1337daysofcode
kind of excited to spend the rest of this month as a full time
#AR
artist - turning software into art that you might see at museums... of the future! ;D
Submitted the starchild spinoff of
#Napkinmatic3D
- "Toy Ripper" to the
#AppStore
... Will the name or nature of the app get me rejected? Will my current super indie state survive server costs?
tldr; as a Taiwanese-American with avoidant complex
An API used for the viral app demo I made is LIVE on
@producthunt
- help vote
@MeshyAI
... currently this
#AI3D
ecosystem provider is
#2
- just a few votes behind a $500M+ funded company!!!
Missed an updated Shap-E from
#OpenAI
#devday
releases - so here's an
#AI3D
app demo using just a mix of
#opensourceAI
models
Napkinmatic3D "Toy Ripper mode" lets you rip any product image into 3D printable parts.
#VisionPro
fits in a
#Hololens2
case
Throwback - so I brought my HoloLens to Apple VisionPro Dev Lab and they made me check it into the locker (I didn't want it taped) / Mac only in the labs
Crashed replicate hackathon on Sunday - hacked out (for fun x $5k win) a card deck ai creator app that lets you mod RWS Tarot as AR markers via controlnet-SDXL
also met
@0xSigil
and donated
@InaCentaur
voice to his cog xtts
At
#AWE2019
we unveiled a new kind of
#AR
musical instrument that turns traditional Chinese brush paintings into 3D "puzzle-note-pieces" to be push-played "back in place":
#NewMedia
#Art
What if you can create a 3D Vision Pro volume app by simply chatting with it?
VolumeMatic Pro lets anyone create 3D interactive games for visionOS.
This is not SORA and not discord. It's a new app I'm building - evolution of AI3D projects I've demo'd at
#SIGGRAPH
and beyond
Old Napkinmatic3D app new use case… turn your Lego creations into the real thing!
Here we turn a Lego bird to a LCM realistic one to a 3D AR model - live iPhone demo - birdy approved by
@corgi_cam
So... I think I'm going to officially retire from
#hackathons
. They are basically technological
#EchoChambers
where even winning
#techcrunch
grand prize has no reach to whom it matters.
Finally added
#GPT4
vision to my Napkinmatic app ... image to code - turn your napkin sketch into a working website - on a spatial computing browser surface!
What's a good one shot image to working web design prompt?
@corgi_cam
friendly
#rainyday
#agi
#hackathon
at home
Should I do another
#AppADay
hack spree for
#VisionPro
#MixedImmersive
#VisionOS
apps only ofc. No boring windowed apps for spatial computing.
Because I like making rare things, esp with an unfair advantage building 3D
#AR
apps for the past 15+ years
DMs open for Requests.
Generally only posting some of the things I make - and only letting myself on social media twice a week max. DMs are checked but replies may not be seen till next week.
I promise I am not a data hoarder.
The biggest glomp is from visionOS derived data dev junk (and then ollama and docker)
Also controlnet-SDXL does intriguing renditions of
#GrandPerspective
into storybook hoarding
i haven't logged into stackoverflow for months or years and suddenly i'm in the top 2% 🤣 - this is my secret account where i used to ask all of my dumb questions before LLMs could rtfm for me