ChatGPT exceeds 1.5 billion generated images per week

OpenAI reports ChatGPT now generates over 1.5 billion images weekly, a 50% surge driven by the launch of Images 2.0 with improved text and consistency.

OpenAI highlights a usage milestone: over 1.5 billion images are now generated weekly in ChatGPT, a volume increase of over 50% in the two weeks following the launch of Images 2.0. This figure is detailed in the nineteenth episode of their in-house podcast, where Product Lead Adele Li and Researcher Kenji Hata, interviewed by Andrew Mayne, retrace the journey since DALL·E's beginnings.

Images 2.0 emphasizes several advancements: improved text rendering, including in multiple languages, enhanced photorealism, real-world knowledge, flexible aspect ratios and resolutions, as well as character consistency from image to image. The model also follows long and detailed instructions, exceeding a thousand words, to produce infographics, PDFs, or educational materials, and handles 360-degree images for more productive uses.

Regarding trends, the discussion highlights virality, a certain acceptance of imperfection, and the importance of precise prompts. Looking ahead, OpenAI mentions creative agents and deeper integration with Codex, which would bring image generation and code closer together.