DmitryUlyanovML's profile picture. Founder @ http://avaturn.me, PhD

Dmitry Ulyanov

@DmitryUlyanovML

Founder @ http://avaturn.me, PhD

Dmitry Ulyanov reposted

We just launched AI avatars 9X faster and 4X more expressive. Here’s why they feel truly human—and why you can’t go back. 👇 [New Avatars Availiable Starting Today: producthunt.com/products/avatu…] Three years ago, we launched our first avatar solution, and we were humbled that…


Great work!

Happy to share our new paper on hair reconstruction that will appear at #ICCV2023, congrats to @ness_pirs and colleagues! We produce a detailed and personalized strand-based hairstyle from the monocular video. Project page: samsunglabs.github.io/NeuralHaircut More details in the thread 🧵



The most requested feature is finally out!

After 4 months in beta Avaturn avatars with ARKit blendshapes support released! We are live on Product Hunt! Support us at producthunt.com/posts/avaturn We also added TONS of new Features (see detailed description in the comment below) Shoutout to @ChrisMessina for hunting us, to…



Removing hair took us longer than adding hair.

Bald&Bold update! 😎 You've been asking and they're finally here: Bald Haircuts! Bring your unique type of the classic sharp look👍 #aigenerated #ainews #avaturn



Dmitry Ulyanov reposted

🔥 #GoogleIO: Google unveils face tracking library for the web, compatible with ARKit blendshapes! 😮 💥 And we already bundled it into Avaturn! 🚀 Now you can create a life-like avatar with a selfie and animate it right in the browser. 💡 Want to try it out? Share and…


So we are finally launching Avaturn. 🚀Well, we actually already posted this a month ago and we managed to drop a database on backend in 5 minutes after the launch. True story, but finally it's 100% live and will not crash again. Support us by sharing!

🚀Announcing the release of next-gen avatar platform, avaturn.me🚀 It’s now a matter of a few seconds to get your realistic, customizable 3D avatar from a 2d photo using Avaturn’s new generative AI. Developer? Integrate into your game or app, in 15 min. for free.



The best part is when people are using what you've been working on for a long time :)

Here's the creepiest video I've recorded. I suppose you've done your share of having fun with the scanned avatars. If not, it works quite well. Model looks good and it's like 17k tris so VRC won't complain.



Excited about this! UE team solved the automation of 'scan geometry -> metahuman geometry', and we solved the automation of 'scan texture -> metahuman texture' for our avatars! See full video at youtu.be/BK-ayKR8NlQ

Now you can turn yourself into #Metahuman with #in3D App! We go beyond «Mesh to Metahuman» by @UnrealEngine by preserving the facial texture so that Metahuman really looks like you. Full video explaining the process: youtu.be/BK-ayKR8NlQ #3davatar #epicgames #metaverse #UE5



LaMa: Resolution-robust Large Mask Inpainting with Fourier Convolutions Inpainting large regions on HD photos! 🌐 Code and paper: github.com/saic-mdal/lama 💻 Web app by @cyrildiagne: cleanup.pictures


We launched our @in3D_io app on Product Hunt today! Come say hi if you are PH regular ;) 🌐producthunt.com/posts/in3d


At @in3d_io we managed to get a runner-up at @NianticEng AR competition. It's fun to run around the place in AR with your character. It certainly adds some flavour over playing a predesigned char. And it was easy to build thanks to ARDK framework!

Shrink yourself down and outwit your environment! in3D by @in3D_io and Likevr, along with curiosity who inspired the project, is a game that uses Meshing to drop you into your surroundings in this platformer. #ardk #ar #lightship #meshing #niantic youtube.com/watch?v=h1qgVR…

NianticEng's tweet card. in3D Niantic ARDK Demo Challenge

youtube.com

YouTube

in3D Niantic ARDK Demo Challenge



Dmitry Ulyanov reposted

Just pushed an update that drastically improves the skeleton. Now it's fully Mixamo compatible and you can plug-in in3D avatars into any game or environment that works with Mixamo rig.

avaturn_me's tweet image. Just pushed an update that drastically improves the skeleton. Now it's fully Mixamo compatible and you can plug-in in3D avatars into any game or environment that works with Mixamo rig.
avaturn_me's tweet image. Just pushed an update that drastically improves the skeleton. Now it's fully Mixamo compatible and you can plug-in in3D avatars into any game or environment that works with Mixamo rig.
avaturn_me's tweet image. Just pushed an update that drastically improves the skeleton. Now it's fully Mixamo compatible and you can plug-in in3D avatars into any game or environment that works with Mixamo rig.

Dmitry Ulyanov reposted

Prompt: "a cold landscape by Thomas Kinkade" (CLIP guided diffusion)

RiversHaveWings's tweet image. Prompt: "a cold landscape by Thomas Kinkade" (CLIP guided diffusion)
RiversHaveWings's tweet image. Prompt: "a cold landscape by Thomas Kinkade" (CLIP guided diffusion)
RiversHaveWings's tweet image. Prompt: "a cold landscape by Thomas Kinkade" (CLIP guided diffusion)
RiversHaveWings's tweet image. Prompt: "a cold landscape by Thomas Kinkade" (CLIP guided diffusion)

This was my first attempt to try-on clothes on the avatar. Created avatar by capturing front and back point cloud with Kinect and aligning them by hand. Simulated cloth in CLO3D. The second image is how it looks now at @in3D_io

DmitryUlyanovML's tweet image. This was my first attempt to try-on clothes on the avatar. Created avatar by capturing front and back point cloud with Kinect and aligning them by hand. Simulated cloth in CLO3D. The second image is how it looks now at @in3D_io
DmitryUlyanovML's tweet image. This was my first attempt to try-on clothes on the avatar. Created avatar by capturing front and back point cloud with Kinect and aligning them by hand. Simulated cloth in CLO3D. The second image is how it looks now at @in3D_io

Wow, finally native VScode support for GitHub!

🤫 New shortcut: Press . on any GitHub repo.



🚀Today we are launching a 3D virtual fitting room with Replicant Fashion house. 30+ designers, 60+ looks. Try them on your digital copy! 🌐 replicant.fashion/digitaltwin (on desktop) 🌐 apple.co/3xCAKDE (iPhone)


Great work, let's see what will win eventually -- end2end neural rendering or deep 3D modeling \w deep physics. Most likely there won't be a fight though and everything will converge to deep 3D + neural rendering as postprocessing

Beyond excited to share our @CVPR'21 paper for learning-based 3D garments. We learn a generative model for 3D garments that directly outputs deformations that *do not collide* with the underlying body. 👗👕 mslab.es/projects/SelfS… A thread on how it works (& cool GIFs) ⬇️ (1/6)



This is insane. The image quality is almost flawless and the network captures even complex concepts as mirrors and reflections.

AI, please give me an emoji of a baby penguin wearing a blue hat, red gloves, green shirt, and yellow pants openai.com/blog/dall-e/

woj_zaremba's tweet image. AI, please give me an emoji of a baby penguin wearing a blue hat, red gloves, green shirt, and yellow pants

openai.com/blog/dall-e/


Cool neural rendering work by @ASevastopolsky from Samsung AI. Creates a 3D portrait of human head that can be relit from a video captured with smartphone. Check out the project page! 🌐 saic-violet.github.io/relightable-po… 🎬 youtube.com/watch?v=qQnNal… 📝 arxiv.org/abs/2012.09963


D'oh, I don't know how to write code anymore

DmitryUlyanovML's tweet image. D'oh, I don't know how to write code anymore

Loading...

Something went wrong.


Something went wrong.