Chinese artificial intelligence (AI) startup Zhipu on Friday unveiled a video generation model, the latest sign of homegrown tech companies gaining dominance in the AI video space.
Ying's text-to-video model accepts both text and image prompts and generates a six-second video clip in about 30 seconds. Users can fine-tune the results with style options like 3D animation, cinematic or oil painting, as well as emotional themes like tension, vitality or loneliness.
The service, accessible through Zhipu AI's chatbot ChatGLM's official website and mobile app, is now available immediately with unlimited availability for all users, although the free version will experience longer waiting times during peak hours, the company said at a launch event in Beijing on Friday.
Ying's launch comes two days after a similar move by short-video app Kuaishou, a rival to ByteDance's Douyin, the Chinese version of TikTok, and shows the Chinese tech company is challenging industry leader OpenAI in the field of video generation.
On Wednesday, Kuaishou opened up its hugely popular Kling video model for more extensive testing, with each customer now able to generate six videos per day.
Kling offers annual paid plans that allow you to generate up to 60 and 800 videos per month for 396 RMB (US$54.63) and 3,996 RMB, respectively.
Meanwhile, San Francisco-based OpenAI, which launched Sora in February and has become a pioneer in AI video generation, has yet to release its model to the public.
Asked about a release date for Sora, Aditya Ramesh, a key member of the development team, said OpenAI wants to ensure that the model is not used to generate or spread misinformation.
The technology behind Ying is a home-grown text-to-video model called CogVideoX, which is similar to the diffusion transformer (DiT) architecture used by OpenAI's Sora and enables faster video generation due to improved inference speed, Zhipu CEO Zhang Peng said on Friday, adding that the company took some inspiration from Sora's algorithm design.
OpenAI has not yet released Sora to the public, but the company has published technical details about how it works.
Zhang also said that Zhipu is working on launching a new version of its video model that can generate longer videos with higher resolution.
