OpenAI Releases Images 2.6 with GIF Generation, Reference Images, and SVG API Support
Frontier AI models · AI creation
OpenAI has launched Images 2.6 across ChatGPT, Codex, and its API. The update introduces GIF generation from prompts or images, reusable reference images, and dual models—Sunburst and Flare. The API also gains SVG output, image markup, and complexity controls. While OpenAI claims a 50% speed increase and 6-second generation under low quality, independent benchmarks were not provided.
OpenAI released its Images 2.6 model on November 4, rolling it out across ChatGPT, Codex, and developer APIs. ChatGPT users can now convert text prompts or images into GIFs with adjustable speed and frame editing, alongside saving reusable reference images for characters. On the API side, Images 2.6 offers GPT-Image-2.6 Sunburst and Flare, introducing image markup, SVG export, and a complexity parameter to govern visual planning. OpenAI claims generation speeds are up to 50% faster, with Flare generating images in as few as six seconds on low-quality settings; however, test conditions were not detailed and remain unverified by third parties. OpenAI's system card notes that the safety filter achieved a 77.7% recall rate on adversarial GIFs, acknowledging ongoing challenges in detecting motion-based violations.
Key takeaways
- Images 2.6 enables GIF generation from text or images with frame editing and adds reusable reference image storage in ChatGPT.
- The API introduces Sunburst for consistency and Flare for speed, maintaining identical per-token pricing to the previous generation.
- New API capabilities include SVG vector output, image markup, and complexity controls to balance planning effort and speed.
- Safety protections retain C2PA metadata and SynthID watermarking; GIF safety filters demonstrated a 77.7% moderation recall rate.
Source：INSIDE Read the original ↗
Prepared, checked and automatically published by AI; not individually reviewed by a human. Translation checked by AI.