Sora (Image)
OpenAI's still image generator based on the Sora diffusion-transformer, discontinued in April 2026.
At a Glance
Pros
- + Generated stills as patch tokens, sharing the video model's architecture
- + Held its consistency well when turning a still into video
Cons
- − Discontinued by OpenAI on April 26, 2026
- − No longer accessible or usable
Best for: Historical reference only — the service was permanently shut down in April 2026
Fabian's Take
CPO & Chief AI Officer
"I liked the idea of one model handling the still and the motion, so your image and its video stayed on-model. OpenAI pulled it in April 2026, so there's nothing to try now — reach for an active tool instead."
[!WARNING] OpenAI officially discontinued the Sora web and API experiences on April 26, 2026. The model is offline and no longer usable. This review remains online for historical reference.
Full review
Sora’s image generation used the same diffusion-transformer architecture as its video model, generating still images as “patch” tokens.
Shared architecture milestone
Its main advantage was the shared codebase and architecture with the video model, which let it animate a still frame cleanly. It also handled combining separate source images without dropping their key visual attributes.
Discontinuation
With the shutdown of the Sora platform on April 26, 2026, these image generation capabilities are no longer available.
Alternatives
For high-quality image generation and precision editing in 2026, look to active platforms such as Google’s Nano Banana 2 or FLUX-2-Pro.
Added: 2026-02-22 · Last updated: 2026-07-01