// news · multimodal · frontier-models2026-05-28source: openai / techcrunch / theverge

OpenAI Sora 2 Pro ships 2-minute clip generation — extended-duration video model launches with per-shot continuity controls and integrated audio synthesis

OpenAI launched Sora 2 Pro on May 28 with 2-minute clip generation, per-shot continuity controls, and integrated audio synthesis. The extended-duration release shifts the AI-video frontier from the 30-second-to-1-minute generation ceiling that prior models hit toward the multi-shot narrative storytelling that production teams have been waiting for. Pricing and tier availability route through the existing Sora API and the new ChatGPT Pro entitlement.

The duration extension is the substantive capability piece. Sora 2 — released in late 2025 — established the 1-minute clip generation ceiling with the per-clip continuity and motion-coherence improvements that made the model the leading text-to-video tool through early 2026. Sora 2 Pro doubles the clip ceiling to 2 minutes and introduces the per-shot continuity control surface that lets production teams chain shots inside a single clip while maintaining character, lighting, and setting continuity. The integrated audio synthesis — dialogue, ambient sound, and synchronized score generated as part of the same generation pipeline — closes the audio-track gap that prior models required ex-post audio dubbing to address.

The competitive context is the AI-video quartet that the field has converged around: Sora 2 Pro, Google's Veo 3.1, Kling 3, and Seedance 2.0. Each model leads on different axes; the per-shot continuity controls and integrated audio synthesis are the Sora 2 Pro differentiators. Runway's Gen-5 physics-aware release the same day extends the multimodal frontier on the physics-realism axis, with the two releases competing on adjacent capability surfaces. For production teams, the practical decision is no longer about whether AI-video tooling is production-ready — it is about which model's specific capability profile matches the project. The differentiation surface is now narrow enough that vertical-specific tooling is emerging on top of each frontier model.

See our analysis →

OpenAI — Sora 2 Pro launch May 28 2026 2-minute clip generation → · TechCrunch — OpenAI Sora 2 Pro extended duration video model → · The Verge — Sora 2 Pro per-shot continuity integrated audio →