EXPLORINEER
en

JP — ITmedia

OpenAI Announces "Images 2.6": 50% Faster Generation, GIF Animations Now Possible

AI2 min read

OpenAI has released ChatGPT Images 2.6, a new AI image tool that makes pictures about 50% faster and can now create GIF animations. It is the successor to Images 2.5, which came out in September.

What happened

On November 4 (U.S. time), OpenAI announced ChatGPT Images 2.6. It makes images faster and edits them more accurately. It also adds two new features: GIF animations, and saved reference photos that you can reuse in other chats. Images 2.6 is available to all ChatGPT and Codex plans worldwide, including the free plan. The GIF feature is still rolling out.

Key numbers and facts

  • Image generation speed is up 50% compared with Images 2.5.
  • The API has two models: GPT-Image-2.6 Flare (fast, at least 6 seconds per image at low quality) and GPT-Image-2.6 Sunburst (higher quality).
  • API prices stay the same as Images 2.5: $8 per million image input tokens, $30 per million image output tokens, and $5 per million text input tokens.
  • OpenAI says ChatGPT, Codex, and the API make about 4 billion images per week, or more than 500 million per day on average.

What is new, in simple terms

Images 2.5 could change one part of a picture and keep the rest the same. Images 2.6 does this better. If you change a character's background, the clothes and small details stay the same. This helps keep characters, products, and brand looks consistent after many edits. The new saved reference photos work like a named photo library. For example, you can save a photo of your dog as "Pepper" and then type "@Pepper watercolor painting" in a new chat, without uploading the photo again.

Why it matters

The GIF feature lets people turn a simple sketch into a waving character or a looping animation. Companies could use this for mascots, product animations, and social media reactions. OpenAI also added a safety check for harmful movement, which still-image checks can miss. It is also adding Google DeepMind's SynthID watermark, alongside the existing C2PA metadata, so people can check where an image came from.

What to watch

  • When the GIF feature finishes rolling out to all users.
  • How well the saved reference photos and consistent editing work in real use.
  • Whether SynthID and C2PA help people tell AI images from real ones.

Glossary

GIF
A short, looping animated picture file.
API
A way for software developers to connect their own apps to an AI model.
SynthID
An invisible watermark from Google DeepMind that marks images as AI-made.
C2PA
A standard that attaches information about where a digital image came from.
Codex
An OpenAI tool that helps people write and work with computer code.

Ad

Add 150 words. Learn 10 a day.

Dicter word books send only the words you learn to your review notes.

More in this topic

AIThe Register

Building AI factories at enterprise scale

A sponsored feature describes how enterprises are increasingly building their own AI infrastructure, driven by data control needs in regulated sectors and the rise of agentic AI. ASUS has unveiled a three-layer Enterprise Agentic AI Solution, announced at NVIDIA's GTC Berlin event, spanning AI servers, a software stack, and a services hub. The company says agentic workloads are increasingly CPU-dependent, which creates bottlenecks in conventional GPU-focused servers.

AIZDNet Korea

LG Shares AI Safety Innovation Cases from 16 Affiliates at Magok Conference

LG held its Safety, Health and Environment Conference, "LG SHEC," at LG Science Park in Magok, Seoul, on the 4th and 5th, with the theme "Drive the Shift." Safety and environment managers from 16 affiliates, about 350 people, attended, and 15 of the 21 innovation cases shared used AI. Examples included LG Electronics' AI-linked CCTV control system, LG Display's AI-based musculoskeletal risk assessment, LG Chem's AI agent for process safety management, and LG U+'s AI risk assessment app.

AIVnExpress

Vietnam Artificial Intelligence Festival 2026 Opens Today

The Vietnam Artificial Intelligence Festival 2026 opens at 2 p.m. on November 5 at 144 Xuan Thuy, Cau Giay, Hanoi, with managers, experts and businesses discussing AI. Organized under the direction of the Ministry of Science and Technology, with VnExpress coordinating and Vietnam National University, the event features speakers from Meta, FPT Smart Cloud and MoMo, and the Vietnam AI Awards 2026. Its theme is "Agentic AI & The Future of Learning."

AITechCrunch

Apple is reportedly cutting budgets for its internal AI usage

Bloomberg reported that Apple is cutting the daily amount its engineering teams can spend on AI, lowering the cap from $300 to $200 in October, with some allowances expected to fall to $100 a day. Apple is also reportedly weighing a daily limit as low as $25 for staff outside engineering. The move comes as companies reconsider the costs of AI tools after a short-lived push to maximize token usage.