TikTok’s parent company, ByteDance, has unveiled a new AI tool that can turn a single photo into a realistic video. Named OmniHuman-1, this tool can make people in images move, talk and even play instruments. ByteDance claims it outperforms other AI tools by creating lifelike videos with minimal input, primarily using just audio.
Works with Any Image Size
OmniHuman-1 isn’t limited to facial expressions; it works with portraits, half-body, and full-body images and creates smooth, high-quality movements. Other AI models usually animate faces, but this tool makes the whole body move naturally.
Read more: UNESCO Dedicates World Education Day to Artificial Intelligence
More than Just Talking Faces
Unlike most AI video generators, OmniHuman-1 goes beyond simple mouth movements. It can create hand gestures and body movements and even show different angles of a person. Researchers have shared sample videos in which it brings both people and animals to life with fluid motion.
Albert Einstein in Motion
One striking example is a black-and-white video of Albert Einstein talking and moving his hands in front of a blackboard. The AI generates realistic facial expressions and hand gestures, making the video look incredibly real.
View this post on Instagram
Trained on Thousands of Hours of Video
ByteDance says it trained OmniHuman-1 using over 18,700 hours of human videos. The AI learned to generate natural movements using inputs like text, audio and physical poses. Researchers believe this method helps it perform better than similar tools.
While OmniHuman-1 isn’t the first AI that turns images into videos, ByteDance may have an edge since it likely used TikTok’s massive video library for training. The company claims this AI produces smoother, more realistic results than its competitors. Only time will tell if it really does.
Stay tuned to Brandsynario for the latest news and updates.