1. What was announced
OpenAI announced ChatGPT Images 2.6, a new version of its image model, on November 4, 2026 (U.S. time). It succeeds Images 2.5 from September, and is said to generate images 50% faster. It is available from the same day in all ChatGPT and Codex plans, including the free plan, though GIF creation is being rolled out gradually (ITmedia NEWS). The system card published the same day describes the update as better editing consistency, detail preservation and response to sketches and image annotations. It adds GIF generation and photo markup in ChatGPT and the API, plus SVG output in the API (OpenAI system card).
GIFs can be requested as looping animations, and a hand-drawn sketch can be turned into, for example, a character waving. Intended uses include company mascots, product animations and social-media reaction images (ITmedia NEWS). Business+IT also describes turning product photos into looping clips and using GIFs as animated storyboards before video production (Business+IT). For the API there are two models: GPT-Image-2.6 Flare, which can generate an image in as little as 6 seconds at low quality, and GPT-Image-2.6 Sunburst, aimed at finer detail and consistency across edits at medium to high quality. ITmedia NEWS reports that API pricing is unchanged from Images 2.5, at $30 per million image output tokens (ITmedia NEWS).
OpenAI says that images generated across ChatGPT, Codex and the API now total about 4 billion a week worldwide, more than 500 million a day on average (Impress Watch).
2. Provenance and safeguards: what to know before business use
According to the system card, images from ChatGPT, Codex and the API will carry Google DeepMind's SynthID watermark in addition to the existing C2PA metadata. The invisible watermark is meant to complement C2PA, and OpenAI says a public verification tool lets people check supported files. For GIFs it adds dedicated safeguards for problems that do not show up in still images, and blocks some content categories more broadly than for stills. At the same time, OpenAI writes that its GIF moderation recall on a set of adversarial test GIFs was 77.7% overall, that motion-related violations remain hard to detect, and that the safeguards may not catch everything (OpenAI system card).
Our view We think the real change for small businesses is that product images and promotional animations can be prototyped in house, without waiting for a vendor's quote and delivery date. With API pricing unchanged from Images 2.5, the cost of automating image replacement falls most for online sellers with many items. A company that starts late will wait days for outsourced images while rivals update in house the same day, and the gap shows in how fast listings go live. Prototype one product first and have a staff member check for differences from the real item. If any appear, we would keep the output off product pages and limit it to social-media promotion.
3. What small businesses should check this week
- Put one of your own product photos into Images 2.6, change only the background, and have a staff member check that color, logo and shape match the real item
- Compare your latest vendor quote and lead time for images and banners with what an in-house prototype would take
- Set an internal rule on where AI-generated images may be used (product pages, ads, social media) and whether to label them as AI-generated; see also our AI adoption Q&A
- If you use GIFs, name someone to check before publication that no trademarks, real people or other companies' characters appear
- Decide whether to use ChatGPT plans or the API, and if the API, estimate cost from your monthly image count
Our outlook
Over the next six months to two years, we expect product images and promotional animations to shift from something people make one at a time to something produced in the needed quantity once product data is fed in. In the same period, consumers will increasingly let AI find and compare products for them, and those comparing AIs will weigh machine-readable information such as product name, dimensions, price and stock, not only pictures. As more companies can produce images cheaply in volume, the gap in appearance narrows, and companies with accurate, quickly updated product data become easier to choose. Those that can run image generation and product-data upkeep in the same system will stay on the shortlist for both human and AI comparison. Now is the time to lay that groundwork.
Questions this article answers
What can Images 2.6 do that it could not before?
The version announced on November 4, 2026 adds animated GIF creation, annotation-based edit instructions, and better preservation of people and products across repeated edits, with generation said to be 50% faster. It is available in all ChatGPT and Codex plans, but GIF creation is rolling out gradually.
Can we put AI-generated product images straight onto product pages?
Before posting, have a staff member check that color, logo and shape match the real item. OpenAI says generated images carry C2PA metadata and a SynthID watermark, so the fact that an image is AI-made can be detected later; we recommend setting your posting rules first.
Sources
- OpenAI: ChatGPT Images 2.6 System Card (2026-11-04)deploymentsafety.openai.com
- ITmedia NEWS: OpenAI、「Images 2.6」発表 生成速度50％向上、GIFアニメの生成も可能にitmedia.co.jp
- ビジネス＋IT: 週40億枚の画像生成 OpenAIが「Images 2.6」でGIF作成と「崩れない編集」を投入news.yahoo.co.jp
- Impress Watch: OpenAI、アニメーションGIFを作れる「Images 2.6」news.yahoo.co.jp
Related Q&A
Related services
If you want to try this in your own business
To see how the topics in this article would apply to your own operations, and what it would cost, send us a message through the contact form.