Sora 2

By OpenAI

Model family: Sora

Sora 2 turns direction into coherent, cinematic footage while keeping characters, props, lighting, and materials consistent from frame to frame. You can start with a prompt or a reference frame, block out a shot, and refine into high-quality renders that honor framing, lens cues, pacing, and choreography. The model is tuned for temporal stability and physical plausibility, so motion reads clearly and fine details—faces, hands, textures, typography—hold up under movement and scene changes. Editing happens in place: extend a take, re-time action, swap backgrounds, or inpaint/outpaint regions without restarting the shot, which makes iteration feel like normal post rather than trial and error. Sora 2 also supports multi-shot workflows, maintaining identity and art direction across cuts for sequences that stitch together cleanly. Teams use it for ads, product demos, explainers, social spots, and pre-viz because it balances fast previews with delivery-grade renders and gives directors reliable control over style and camera language.

New Video Gen 3

Released: September 30, 2025