ChatGPT can now make GIFs with the new Images 2.6 model

OpenAI has launched Images 2.6, a new image model that brings GIF creation and saved character references to ChatGPT. It is rolling out across every subscription tier, and generation is about 50% faster than before.

What happened

Images 2.6 is now rolling out in ChatGPT (every subscription tier) and in Codex (OpenAI's coding agent, an AI that completes a task with several steps for you, not just one answer). Developers can also access it through the API. OpenAI says people make about 4 billion images a week across its products.

The headline additions are:

• GIFs: short animated loops generated from characters, mascots, product shots, sketches, or storyboards.
• Saved references: store a picture of yourself, a pet, a character, or a product once, then reuse it across multiple requests without re-uploading each time.
• Speed: about 50% faster, with the API's speed-focused model able to produce an image in as little as six seconds at low quality.
• Complexity setting: controls how much planning the model does before drawing. OpenAI recommends low for quick edits and medium or high for crowded scenes with many constraints.
• For developers: image markup (mark the exact spot on an image instead of describing it in words), SVG export (saves a design as shapes stored as instructions so it stays sharp at any size), and two API models split by use case: GPT-Image-2.6 Flare for speed and layouts like posters, and GPT-Image-2.6 Sunburst for detail and consistent edits.

The saved references feature addresses a common frustration: image models normally do not remember your character or product between requests, so keeping a mascot looking consistent across several scenes meant re-uploading a reference photo every single time.

What it means for your work

• You can create short animated GIFs for social posts, presentations, or client updates directly in ChatGPT, without a separate design tool.
• Saved references mean your brand mascot, product, or headshot can stay visually consistent across a whole set of images in one session.
• Faster generation means less waiting when you are iterating on a design.
• SVG export is useful for logos and icons because the file stays sharp at any size and a designer can still edit it. It is not yet clear how well it handles photo-like images, so start with logos.

What to do now

1. Open ChatGPT and look for the image generation option. The rollout covers all subscription tiers, though timing may vary.
2. Try the GIF feature on a logo, mascot, or product shot you use regularly. Describe the character and ask for a short animated loop.
3. Test saved references by uploading a character or product image and asking ChatGPT to place it in several different scenes. Check whether scene five still looks like scene one.
4. If you need a logo or icon, try the SVG export option so the file scales without losing quality.
5. Use the complexity setting: choose low for a quick tweak and medium or high when your scene has many elements or specific constraints.

The real test for Images 2.6 is consistency: if your character still looks like itself after several edits and scene changes, the saved references feature is doing its job. If not, faster generation alone does not solve the problem of image generation starting from scratch every time.