ByteDance, the tech company behind social media platform TikTok, has introduced a new AI model called OmniHuman-1, capable of generating lifelike and dynamic videos.
ByteDance claims that OmniHuman-1 uses deepfake technology to create videos with incredibly smooth movements that are nearly indistinguishable from real footage. The company boasts that its AI tool outperforms existing video-generating AI tools in terms of quality.
According to ByteDance researchers, OmniHuman-1 requires only a single image or audio sample, such as a voice or singing tone, to generate videos of any length with voices or singing identical to the sample. The AI tool can also edit and modify existing videos to change their original content.
Moreover, OmniHuman-1 allows users to create new videos with various styles, such as animations, cinematic styles, or realistic human-like videos.
Users simply need to specify frame rates, body proportions (full-body, half-body, close-up), and other preferences, and the AI will produce highly realistic videos with accurate movements, lighting, and details.
While ByteDance has not yet officially released OmniHuman-1 to the public, they have shared sample videos generated by the tool.
One 23-second video features Albert Einstein delivering a speech with lifelike facial expressions and lip movements.
Other short videos showcase people giving presentations with gestures and facial expressions that appear real.
These sample videos have impressed both the online community and the tech industry, with many acknowledging them as some of the most realistic fake videos they’ve seen.
ByteDance researchers claim to have built an AI model that blends data from images, videos, and audio to create highly realistic videos, matching facial ratios, body proportions, facial expressions, and actions as closely as possible.
The team says that the AI-generated videos feature natural-looking faces, head movements, and gestures that match the audio and surrounding scenes, enhancing the realism of the videos.
ByteDance has not yet disclosed when OmniHuman-1 will be available to users.
The emergence of OmniHuman-1 highlights the impressive capabilities of AI tools from China but raises concerns about potential misuse for malicious purposes, such as creating fake videos of celebrities to spread misinformation, commit fraud, or even generate fake explicit videos for blackmail.
The race to develop advanced AI is intensifying between the U.S. and China, with both superpowers continuously introducing superior AI tools to outdo each other.
Recently, OpenAI integrated a new feature into ChatGPT, allowing it to complete a report in just a few minutes from vast amounts of data, a task that would take humans hours.
Deepfake is an AI technology that uses image synthesis techniques to overlay one person’s face onto another’s video or image.
Deepfake undergoes a “learning process,” using input data to create models and select appropriate algorithms for continuous processing and learning. After sufficient training, deepfake can seamlessly integrate faces with high accuracy, making them almost imperceptible to the naked eye.
