TikTok creators present an AI tool for making videos

ByteDance, the owner of TikTok, has introduced OmniHuman, a new AI tool for creating realistic videos from a single photo. OmniHuman can make it accurately mimic gestures, facial expressions, and even full-body movements, making images sing, talk, and move.

This was reported by Tech Radar.

How OmniHuman works

Developers trained the model on more than 18,700 hours of video to accurately capture human plasticity. The technology makes it possible to create full-featured videos, not just facial or torso animations. An ordinary photo can be transformed into a video of a performance, a dance, or even the playing of a musical instrument.

With OmniHuman, it is possible to animate not only real photos, but also stylized portraits, creating realistic videos with a high level of detail.

Once this technology is made available to the public, it can immediately gain significant popularity on TikTok. The platform already has an AI video generator called Jimeng, although OmniHuman may be able to attract more users.

Other competitors to OmniHuman include OpenAI Sora, Pika, Runway, Pollo, and Luma Labs Dream Machine.

The new technology can be used for:

• restoring images of historical figures for educational purposes;
• creating digital avatars for social networks and games;
• use in movies to recreate actors of the past.

Currently, OmniHuman remains a research project with impessive capabilities.

As a reminder, ChatGPT presented a new feature called Deep Research, which allows for in-depth research. Artificial intelligence can work autonomously, conducting planning and multi-stage research.

We also wrote that YouTube started testing new features for Premium subscribers. One of the most interesting is the ability to watch videos at up to 4x speed. The company is also testing other experimental features that can improve the user experience.