OpenAI is teasing the Image V2 model.

OpenAI has released a teaser for Sora V2, the highly anticipated next generation of its text-to-video AI model. The preview, shared via a Reddit screenshot, highlights dramatic improvements over the original Sora. Most notably, the model can now generate videos up to 60 seconds long, a threefold increase from V1's 20-second limit. It also produces videos in full 1080p resolution, a significant step up in visual fidelity. The teaser emphasizes enhanced character consistency, suggesting the model can maintain the appearance of subjects throughout a scene more reliably, a common challenge for AI video.

Beyond length and resolution, Sora V2 appears to have made substantial progress in simulating real-world physics and object permanence. This means objects and characters are expected to interact with their environment in more believable ways, and elements that move out of frame can be remembered and rendered consistently upon return. These technical advancements aim to produce more coherent and narratively complex videos. The development signals OpenAI's commitment to leading the generative video space, directly competing with other emerging models from companies like Runway and Pika Labs.

The release of such a teaser is a strategic move, likely aimed at maintaining developer and public interest while the model undergoes further testing and safety evaluations. It sets a new benchmark for AI-generated video, pushing the boundaries of what's possible in automated content creation. For professionals, this evolution means the potential for creating longer-form promotional content, educational explainers, and storyboarding with a single AI tool, drastically reducing production time and cost for high-quality video assets.