OpenAI’s Sora 2 lets users insert themselves into AI videos with sound

On Tuesday, Openai announced Sora 2, its second-generation AI synthesis, which can now generate videos in various styles with synchronized dialogue and sound effects, which is the first for the company. Openai also launched a new social application for iOS, which allows users to insert themselves into the video generated by AI through what Openai calls “Kamei”.

Openai demonstrated a new model in a video generated by AI, which depicts the photorealistic version of Openai General Director Sam Altman, talking with a camera in a slightly unnatural voice in fantastic backgrounds, such as a competitive riding of a duck race and a luminous mushroom garden.

As for this voice, the new model can create what Openai calls “complex background sound landscapes, speech and sound effects with a high degree of realism.” In May Google I see 3 He became the first model of video synthesis from a large laboratory of artificial intelligence to generate synchronized sound, as well as a video. Only a few days ago Alibaba released WAN 2.5Open-weight video, which can also generate sound. Now Openai has joined an audio party with Sora 2.

The model also has noticeable improvements to the visual sequence compared to the previous video of the Openai model, and can also follow more complex instructions on several pictures, while maintaining coherence between them. The new model represents what Openai describes as “GPT-3.5 Moment for the video”, comparing it with the ChatGPT breakthrough during the evolution of the text generation models over time.

Sora 2, according to, demonstrates improved physical accuracy compared to the original Sora model from February 2024Since Openai claims that the model can now simulate complex physical movements, such as Olympic gymnastics and triple axles while maintaining realistic physics. Last year, shortly after the launch of Sora 1 Turbo, We saw A few noticeable failures with similar tasks about video generation that Openai claims to be considered with the new model.

“Previous videos are excessively optimistic – they will change objects and deform reality in order to successfully perform a text hint,” Openai wrote in his ad. “For example, if a basketball player misses a shot, the ball can spontaneously teleport to the hoop. In Sora 2, if a basketball player misses a shot, he will retreat from the rear side. ”