AI tool OmniHuman-1 that creates realistic videos, introduced by TikTok's parent company ByteDance

OmniHuman-1: OmniHuman-1 can create videos with realistic human motion from a single image and audio sample. It also supports the image aspect ratio. Company researchers say that in this development, the AI ​​tool has been trained on 19000 hours of video footage. It has not been made publicly available yet. At present, its release timeline is also not clear.

Thu, 06 Feb 2025 07:53 PM (IST)
AI tool OmniHuman-1 that creates realistic videos, introduced by TikTok's parent company ByteDance
AI tool OmniHuman-1 that creates realistic videos, introduced by TikTok's parent company ByteDance

ByteDance, which owns TikTok, recently forayed into AI. The corporation designed an AI platform, OmniHuman1. With its AI platform, one can produce lifelike animated videos with a single photograph. All information regarding it in the study report and samples of several clips have been posted by the corporation.

Although there are many video creation tools, it claims to produce the best videos. It is not yet accessible to its users at the current time.

OmniHuman-1 can generate videos with real human motion out of a single picture and an audio clip. It even accommodates the image proportion. Regardless of whether it is a full-face, a half-body, or a full-body picture.

The company's researchers say that in such development, the AI tool has been trained with 19,000 hours of video recordings. It is argued that it performs even better than current techniques with no high-quality data.

OmniHuman-1 includes multiple input sources, including images, audio, body poses, and textual details. It works well in different scenarios. The team has also shared video clips generated from this tool. In this, the company has shared a video of Albert Einstein, in which there is a video of him giving an explanation in black and white in front of a blackboard. Albert Einstein's video has been prepared with this tool.

This footage looks real, it seems as if the person is actually doing that work. Facial expressions and body gestures seem accurate in it. OmniHuman-1 is currently in the research stage. Its release timeline is not clear at the moment. This tool is not yet publicly available. After the tool is available to users, it will compete with AI platforms like OpenAI’s Sora, Runway, and Pika.

ByteDance's AI model has come at a time when the whole world is discussing China's DeepSeek. DeepSeek AI has been released recently. The Chinese startup has developed this platform at a very low cost, due to which it is also very cheap.

Muskan Kumawat Journalist & Content Writer