Chinese tech giant Alibaba is developing Tora, a video generation tool based on Sora. This was reported by SCMP.
The tool is being developed on the Diffusion Transformer (DiT) architecture, which underpins the neural network introduced by OpenAI in February.
The newspaper notes that experts have created a “trajectory-oriented DiT framework for video generation.” This means that the movements generated by Tora precisely follow a given trajectory, replicating the dynamics of the physical world.
“We adapted the OpenSora workflow to transform raw videos into high-quality video-text pairs and used an optical flow estimator to extract the trajectory,” they said.
It is unknown when the new tool will be available to the general public. Earlier, other Chinese companies introduced their own AI-based video generators.
On July 24, Kuaishou’s Kling model became available worldwide. Later, the startup Zhipu presented its solution, Ying.
In August, Faceu Technology, owned by Chinese giant ByteDance (owner of TikTok), launched the Jimeng AI neural network for generating videos from prompts.
Back in February, OpenAI unveiled a new generative AI model, Sora, which allows for text-to-video conversion.
In June, Toys “R” Us faced a wave of criticism after releasing an advertisement entirely created using OpenAI’s tool.