Sora‑2: OpenAI’s video API
OpenAI published a code‑first guide comparing Sora‑2 and Sora‑2‑Pro, positioning Sora as a rapid prototyping tool that generates short, high‑quality videos from text prompts while flagging limits like character fidelity and occasional audio/visual artifacts. The docs frame Sora as useful for storyboarding and previsualization, not final polish. ( )
POST /videos returns an asynchronous job object (job_id) that clients poll or receive via webhooks, and completed clips can be fetched as MP4 with GET /videos/{video_id}/content. (developers.openai.com) The API exposes programmatic video-edit capabilities — reuse character assets across generations, apply targeted edits, continue clips with “video extensions,” and submit large offline render queues through a Batch API while downloading finished videos plus supporting assets. (developers.openai.com) OpenAI bills Sora per second: the sora-2 baseline is listed at $0.10/second, while sora-2-pro prices scale by resolution (example listed rates: $0.30/s for 720×1280, $0.50/s for 1792×1024, and $0.70/s for 1920×1080). (developers.openai.com) Resolution and format tiers differ by model: sora-2 is capped at ~1280×720 outputs, whereas sora-2-pro publishes higher-resolution landscape and portrait options up to 1792×1024 and 1920×1080. (neowin.net) API access has been rolled out on an invite/org‑verification basis via platform.openai.com, with model snapshots (e.g., sora-2-pro-2025-10-06) available to lock behavior and rate limits applied per usage tier. (sora2ai.run) Notable operational caveats recorded by third‑party coverage: the API currently does not support image‑to‑video for real people (the “cameo” capability) and, unlike the Sora iOS app, the API path does not impose an app watermark on rendered MP4s. (neowin.net)
