OpenAI Images 2.6 Release Notes

Overview of Images 2.6

OpenAI has released Images 2.6, an update to its image generation suite available in ChatGPT, Codex, and the API. The update focuses on performance, introducing a 50% increase in generation speed and new tools for creating animated GIFs and vector graphics. OpenAI reports that its platforms now facilitate the creation of approximately 4 billion images per week.

Performance and API Enhancements

Images 2.6 introduces two distinct models for API users: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. Flare is optimized for speed, capable of generating images in as little as six seconds when set to low quality, while Sunburst is designed for higher-quality output and improved consistency during iterative edits.

New API features include:

  • Complexity Parameter: Developers can now adjust the model's planning effort, allowing for faster, lower-complexity edits or higher-complexity generation for complex scenes.
  • Image Markup: Users can pinpoint specific areas for modification, improving control over the editing process.
  • SVG Export: The API now supports native SVG output, which is achieved by generating a reference raster image and converting it to vector format.

GIF Animation Features

Images 2.6 enables the creation of animated GIFs, allowing users to animate characters or product shots. The process often involves generating a sprite sheet—a grid of sequential frames—which the model then interprets into a looping animation. This feature is intended for use cases such as social media reactions, character storyboards, and UI elements like typing indicators.

Community Reception and Technical Feedback

Community response on platforms like Hacker News has been mixed, with users highlighting both the utility and limitations of the new release:

  • Consistency and Control: While some users appreciate the new editing tools, others report that maintaining consistency across multiple iterations remains a challenge. One user noted, "I want to change a tiny thing and it changes too or not exactly what I want, end up being a long conversation to get what I desire."
  • Technical Implementation: Developers noted that the SVG support functions as a raster-to-vector pipeline rather than native vector generation. Some users pointed out that while the generation is faster, the token usage for certain quality settings has increased, despite pricing remaining stable.
  • Model Quality: Early comparisons on the LMSYS Chatbot Arena and anecdotal testing suggest that while the new model offers speed improvements, some users find the output quality of previous versions, such as 2.5, to be more reliable for specific artistic or technical prompts.

"The API creates a reference image and uses it to produce the SVG. The quality setting controls that reference image. Decode data[0].b64_json as base64 and save the resulting UTF-8 XML bytes with a .svg extension." — @minimaxir

"2.5 is better than 2.6 I've been playing with it for the last hour or so... I've found 2.6-flare specifically to either be too simplistic in its approaches, or create outputs that are inherently broken in other cases." — @jjcm

Sources

Related

  • Dispatch
  • Dispatch
  • Dispatch
  • Dispatch
  • Dispatch