- 14
- 195 320
Flawless AI
United Kingdom
Registrace 24. 08. 2021
Flawless is pioneering the generative AI revolution in filmmaking.
Empowering the world’s greatest storytellers with magical new tools and emerging technologies.
Empowering the world’s greatest storytellers with magical new tools and emerging technologies.
[CVPR '24 Presentation] FaceLift: Semi-supervised 3D Facial Landmark Localization
CVRP 2024 presentation.
We introduce a novel semi-supervised learning approach that learns 3D landmarks by directly lifting (visible) hand-labeled 2D landmarks and ensures better definition alignment, without the need for 3D landmark datasets. To lift 2D landmarks to 3D, we leverage 3D-aware GANs for better multi-view consistency learning and in-the-wild multi-frame videos for robust cross-generalization. Empirical experiments demonstrate that our method not only achieves better definition alignment between 2D-3D landmarks but also outperforms other supervised learning 3D landmark localization methods on both 3DMM labeled and photogrammetric ground truth evaluation datasets.
Paper: arxiv.org/abs/2405.19646
Project Page: davidcferman.github.io/FaceLift/
We introduce a novel semi-supervised learning approach that learns 3D landmarks by directly lifting (visible) hand-labeled 2D landmarks and ensures better definition alignment, without the need for 3D landmark datasets. To lift 2D landmarks to 3D, we leverage 3D-aware GANs for better multi-view consistency learning and in-the-wild multi-frame videos for robust cross-generalization. Empirical experiments demonstrate that our method not only achieves better definition alignment between 2D-3D landmarks but also outperforms other supervised learning 3D landmark localization methods on both 3DMM labeled and photogrammetric ground truth evaluation datasets.
Paper: arxiv.org/abs/2405.19646
Project Page: davidcferman.github.io/FaceLift/
zhlédnutí: 361
Video
![[SIGGRAPH '24] HQ3DAvatar: High-quality Implicit 3D Head Avatar](http://i.ytimg.com/vi/69L2i2-veV4/mqdefault.jpg)
![[SIGGRAPH '24] HQ3DAvatar: High-quality Implicit 3D Head Avatar](/img/tr.png)
[SIGGRAPH '24] HQ3DAvatar: High-quality Implicit 3D Head Avatar
zhlédnutí 451Před 2 měsíci
SIGGRAPH 2024 paper. We present a novel approach to building highly photorealistic digital head avatars. Our method learns a canonical space via an implicit function parameterized by a neural network. It leverages multiresolution hash encoding in the learned feature space, allowing for high-quality, faster training and high-resolution rendering. At test time, our method is driven by a monocular...
![[CVPR '24] FaceLift: Semi-supervised 3D Facial Landmark Localization](http://i.ytimg.com/vi/coUos_F4rmU/mqdefault.jpg)
[CVPR '24] FaceLift: Semi-supervised 3D Facial Landmark Localization
zhlédnutí 151Před 2 měsíci
CVPR 2024 paper. We introduce a novel semi-supervised learning approach that learns 3D landmarks by directly lifting (visible) hand-labeled 2D landmarks and ensures better definition alignment, without the need for 3D landmark datasets. To lift 2D landmarks to 3D, we leverage 3D-aware GANs for better multi-view consistency learning and in-the-wild multi-frame videos for robust cross-generalizat...
![[CVPR 2023 Presentation] Few-Shot Geometry-Aware Keypoint Localization](http://i.ytimg.com/vi/SurF_6SYcx4/mqdefault.jpg)
[CVPR 2023 Presentation] Few-Shot Geometry-Aware Keypoint Localization
zhlédnutí 1,6KPřed rokem
CVPR 2023 presentation We present a novel formulation that learns to localize semantically consistent keypoint definitions, even for occluded regions, for varying object categories. We use a few user-labeled 2D input images, extended via self-supervision using a larger unlabeled dataset. We also introduce 3D geometry-aware constraints to uplift keypoints, achieving more accurate 2D localization...
![[CVPR 2023 Presentation] Implicit Neural Head Synthesis via Controllable Local Deformation Fields](http://i.ytimg.com/vi/_Mqp4zoJJto/mqdefault.jpg)
[CVPR 2023 Presentation] Implicit Neural Head Synthesis via Controllable Local Deformation Fields
zhlédnutí 478Před rokem
CVRP 2023 presentation. We propose an approach for implicit neural head synthesis at high fidelity from 2D monocular video input. Our approach models multiple implicit deformation fields with local semantic control via 3DMM-based parameters and representative facial landmarks. We promote sparsity of each field via a new local control loss and attention mask mechanism. Our formulation renders sh...
![[CVPR '23] Implicit Neural Head Synthesis via Controllable Local Deformation Fields](http://i.ytimg.com/vi/Bg87Fg2VYfw/mqdefault.jpg)
[CVPR '23] Implicit Neural Head Synthesis via Controllable Local Deformation Fields
zhlédnutí 2,3KPřed rokem
CVRP 2023 paper. We propose an approach for implicit neural head synthesis at high fidelity from 2D monocular video input. Our approach models multiple implicit deformation fields with local semantic control via 3DMM-based parameters and representative facial landmarks. We promote sparsity of each field via a new local control loss and attention mask mechanism. Our formulation renders sharper l...
![[CVPR '23] Few-Shot Geometry-Aware Keypoint Localization](http://i.ytimg.com/vi/590XrUqpuR8/mqdefault.jpg)
[CVPR '23] Few-Shot Geometry-Aware Keypoint Localization
zhlédnutí 1,8KPřed rokem
CVPR 2023 paper. We present a novel formulation that learns to localize semantically consistent keypoint definitions, even for occluded regions, for varying object categories. We use a few user-labeled 2D input images, extended via self-supervision using a larger unlabeled dataset. We also introduce 3D geometry-aware constraints to uplift keypoints, achieving more accurate 2D localization. Our ...
How to make a movie work in another language - Voice of America
zhlédnutí 2,1KPřed rokem
How to make a movie work in another language - Voice of America
AI makes actors "speak" foreign languages convincingly - CNN
zhlédnutí 3KPřed rokem
Co-founder/CEO talks to CNN about Flawless
Flawless use of TrueSync on the movie "Fall"
zhlédnutí 154KPřed rokem
Flawless use of TrueSync on the movie "Fall"
"Fall" - Using Flawless technology
zhlédnutí 20KPřed 2 lety
Director Scott Mann and the movies lead actors Grace Caroline Currey and Virginia Gardner talk through using Flawless TrueSync technology in post-production, where we were able to visually change some lines of dialogue throughout the movie to change it from a R rating into PG-13, that helped it reach a much larger audience. The process, done during the final stages of post-production, took less...
How cinema is getting an AI makeover - BBC Click
zhlédnutí 3,9KPřed 2 lety
A few weeks ago the BBC flagship technology series "Click" came to talk to our co-founders and CEOs Scott Mann and Nick Lynes to chat about what we're doing. And here it is...
We Are Hiring
zhlédnutí 560Před 2 lety
We are hiring, Devops, Machine Learning, Software Engineers. Check out www.flawlessai.com/careers for more info.
We Are Flawless
zhlédnutí 4,6KPřed 2 lety
We are filmmakers, we are scientists, we are technologists. We... are Flawless. Creators of TrueSync, the world's first visual dubbing technology using AI.
Can I make it say something else perfectly in English ? And how can I get this? I'm interested
The future is here
It would be cool to see monkey magic translated accurately into English using the original Japanese actors voices and some Kurosawa movies in English with actors like Toshiro Mifunes original voices.
Bruce lee movies in English with his voice
Did Prey from 2022 use this technology? I watched it in the Comanche dub and the audio seemed perfectly synced with the lips except for one or two scenes.
What name is Called Websites 🙏
spanish suck
It seems to flatten the eyebrows too much, when really they stick out further from the face than estimated.
Why don’t they use this?
This is amazing.
Es bueno lo malo puede dobuje en peligro
What name Websites Please plz don't have🙏 ❤
This is probably a type of deepfake btw! There’s technology that can change lip movements to sync with new audio :) Google “Truesync”
La IA es una herramienta y una cosa es que lo haga y habrá un nuevo empleo que será revisión de doblaje IA
I mean ya it’s cool but it’s also really… scary to say the least
We are fucked….
im interested sir...
Yay!! More power to live human performances in theater! 😂
I love the PG13 option.
*Wheres the download link for this programme?*
this is how we should be using AI. can't wait to watch dubbed movies with perfect lip-sync
Trash
How much for month
HERE IS THE MILLION DOLLAW QUESTION: Is this an actual product and if so, how can you buy it? Flawless has been bragging about this for years but it is not available for purchase and their website is useless.
Most likely bought or they were silenced before this ever became a thing. The only way to use AI for dubbing now is through RVC or the character crumbs given on 11labs.
HERE IS THE MILLION DOLLAW QUESTION: Is this an actual product and if so, how can you buy it? Flawless has been bragging about this for years but it is not available for purchase and their website is useless.
Is this tech available to use? I would like to create a perfect dub of the television series Friends into Russian language. Please Respond.
Lol 🤣
do someone actually know anything about this software? is it paid? Wav2lip is so damn bad and outdated
0:08 me when my dad scolds me
0:08 bro she's gonna explode
Oh this is... oh god imagine the consequences? What could go wrong lol
Those FREAKING towers
This is a very complex problem.
Where are the new demos? Outdated!
Is there an open project of that tech on GitHub?
This is an amazing tech with incredible potentials… it’s up to us as a society to make sure that we are taking care of our people with laws and practices.
laws will never catch up. humans need to learn to let people do what they want without controlling what they think and create
She's pretty Virginia Gardner, beautiful 😍
How can you use this app please reply
I wanna watch the no dialogue version. 🤣🤣🤣
No wonder they're going on strike!
But what about animation (Cgi animation,hand drawn animation,anime, etc) when will it ever be possible to change lip sync for that?
to the point ✅. thanks for not creating a 20 minutes video explaining the same thing you did in under 1 minute.
Wow
WOW!
can you make it free for new youtuber?
When Will same voice dubbing Will be possible?
Which program can do that?
Use Flawless AI for The Chosen so the dubbing in different languages is less distracting!
Great for ADR…not so great for the courtroom.
This type of Ai technology opens a new chapter on copyright issues. It is voice and image manipulation of the original actor. A true challenge for Hollywood Unions and Entettainment Law Firms.
I'd like to think it shouldn't be too hard, as long as the talent and the studio are in agreement on what can and can't be done with their likeness... But we got this protest going on right now, so you're right, it's indeed a challenge.