ChatGPT to get Sora video
OpenAI is integrating its Sora video generator directly into ChatGPT, moving the product from text/images to real‑time video creation — a shift that could change how users create and share media inside workflows plans. This pushes multimodal agents toward handling video inputs/outputs and forces product teams to rethink ingestion, moderation, and storage patterns for in‑app generated video.
The Information first reported the internal plan on March 11, 2026, citing people familiar with OpenAI’s product discussions. (theinformation.com) OpenAI disclosed ChatGPT reached about 900 million weekly active users in a February 27, 2026 company update. (techcrunch.com) OpenAI released Sora 2 and an invite‑only Sora app on September 30, 2025, framing the launch around enhanced physics and synchronized audio/video. (openai.com) App market telemetry from Appfigures showed Sora’s installs dropped 45% month‑over‑month to roughly 1.2 million in January 2026. (techcrunch.com) Reporting flagged that folding video generation into ChatGPT would materially raise inference and delivery costs for the flagship product. (theinformation.com) OpenAI’s Sora launch notes require visible watermarks and invisible provenance signals on outputs as a core safety control introduced at the Sora 2 debut on September 30, 2025. (openai.com) Coverage of the proposed ChatGPT integration explicitly called out heightened deep‑fake risk vectors if video outputs are broadly available inside chat flows. (theverge.com) Microsoft documented Sora 2 availability inside Microsoft 365 Copilot’s Create experience with storage options routed to Clipchamp and OneDrive for Business for enterprise governance. (techcommunity.microsoft.com) OpenAI’s infrastructure moves—an October 2025 6‑gigawatt AMD Instinct agreement and a September 2025 strategic partnership to deploy at least 10 gigawatts of NVIDIA systems—establish the multi‑GW capacity being marshalled to absorb video workloads. (openai.com) Combined, those product and infrastructure signals imply engineering teams will need streaming/partial‑render pipelines, per‑tenant quotas and backpressure, and provenance‑first object stores to operationalize short‑form video generation at enterprise scale (inference supported by the Microsoft and OpenAI commitments). (techcommunity.microsoft.com)
