Pinscreen’s advanced AI facial synthesis technology, paGAN, is not only for producing stunningly real digital 3D humans, but can also be used to swap a person’s face with someone else’s in a video at unprecedented fidelity. Unlike existing deep fake software, our solution can ensure high-resolution output, making it suitable for high-end VFX production. We also offer real-time face swapping solution for app filters, live events, TV broadcasting, and online video streaming.
Our HD face replacement technology can replace the face of a person in a video with whoever you like.Our neural rendering workflow ensures AI generated faces that are nearly indistinguishable for real ones, while preserving the likeness of the target person and photorealistic natural facial expressions. We can process faces with video resolutions of up to 1080p.
Given a video sequence with subject A (source), the task is to replace subject A’s face with the one from a subject B (target). All we need are a few minutes of training footages from target subject B and we can generate a perfect face replacement video. Our team will provide you guidance for how to record training videos optimally for subject B.
We support HD video resolution (1080p)
Synthetic faces that are nearly indistinguishable from real ones.
Data collection and training takes only a few days per subject.
One the model is trained we can generate 1-2 frames per second.
Digital Doubles in Visual Effects
Virtual news anchors and lip-sync dubbing
Cybersecurity and Deepfake Detection
Pinscreen’s RT face replacement is a boosted high-fidelity face swapping version that can perform at 30 fps for a face at 512x512 resolution. In addition to real-time performance, we support instant user switching allowing anyone to swap their face with a target subject without training the system as required in existing non real-time deepfake technologies.
Given a live video stream such as a webcam, our system can swap anybody’s face with a target subject instantly. The user’s face is immediately replaced with the one of a target subject and no training is required for each subject. We simply require you to provide us the desired target subject and we will either collect the data for you or provide you with guidelines on how to capture the person you want.
Digital Doubles in Visual Effects
Photorealism: synthetic faces that are superior to all existing face swapping
Pre-Processing: target data collection and training takes only a few days per subject
Real-time performance: we can generate face swaps at 30 frames per second
Instant swapping: the source person does not require to be trained
Pinscreen’s virtual avatars and services can generate humans that are nearly indistinguishable from real ones and our AI-driven technology continues to improve. Many of our results are undetectable by state-of-the-art deepfake detectors. We are building many exciting applications for next generation human computer interaction, but these highly convincing results also have the potential for misuse.
Pinscreen is committed to responsible implementation of its technology to prevent malicious use, such as harassment, disinformation, and invasion of privacy, which is reflected in our company policy and products.
Pinscreen does not develop products that can be harmful or offensive to the general public.
We do not allow end-users to use our technology to generate disinformation and all content created by Pinscreen is screened carefully.
Pinscreen is committed to actively raise awareness on dangers of AI-synthesized media and educate the public on best practices for such disruptive technologies.
We regularly publish our research in top peer-reviewed scientific venues to make our knowledge accessible to the security community, but do not release source code when there is potential misuse of the technology.