ChatGPT can now make a real looping GIF in one step: type what you want (or upload a photo), ask for a GIF, then adjust playback speed and edit individual frames before you save or share. OpenAI launched this with Images 2.6 on November 4, 2026. It is available globally in ChatGPT and Codex on web, desktop, and mobile for all subscription tiers, including Free. This guide is for people who use the ChatGPT app, not for developers first. You will see what to type, how to fix a bad loop, where GIFs are saved, what still only works as a still image, and (in an optional appendix) how the API’s Flare and Sunburst models expose GIF settings if you ever need them.

Everything below was checked on November 5, 2026 (Asia/Shanghai) against OpenAI’s Introducing Images 2.6 announcement, the ChatGPT Images 2.6 System Card, the Help Center article Images in ChatGPT, OpenAI’s image generation API guide (GIF fields and limits), and API pricing. The Verge’s same-day write-up is used only to corroborate the public demo UI (prompt text and speed/frame controls), not as a substitute for OpenAI’s own pages.

Key facts:

• What launched: Images 2.6 in ChatGPT, Codex, and the API, including native GIF generation, faster still-image generation (OpenAI says about 50% faster), and (in the API) Flare/Sunburst models, markup edits, SVG export, and a complexity control (OpenAI announcement).
• Who gets ChatGPT GIFs: “available globally today in ChatGPT and Codex across all subscription tiers on desktop, mobile, and web” (announcement).
• How you start in ChatGPT: ask for a GIF in chat (OpenAI’s demo used “can you make a gif of a cat dancing?”), or upload a photo and ask to animate it; then use playback-speed and frame-by-frame controls (The Verge summarizing OpenAI’s demo).
• Help Center lag: as of this writing the Help Center Images in ChatGPT page still describes Images 2.5 and does not yet document GIF controls. Prefer the November 4 announcement and system card for 2.6/GIF facts; use the Help Center for the shared create/edit/save workflow that still applies.
• Safety: GIF generation adds motion-specific risks; OpenAI added GIF safeguards on top of the Images 2.5 stack (system card).

What Images 2.6 actually changes for you

Before Images 2.6, ChatGPT’s image tools were strong at still pictures: posters, product shots, edits with a selection tool, templates, and (on mobile) sketches. The Help Center still walks through that still-image workflow under the Images 2.5 branding: open Images, describe what you want, edit with Select or chat, then Copy, Save, or Share (Images in ChatGPT).

Images 2.6 keeps that foundation and adds motion. OpenAI’s product post says you can “bring motion to your images with GIFs — from animating a brand mascot or character to turning a product shot into an eye-catching loop,” and that Images 2.6 also generates images about 50% faster than before (announcement). OpenAI also says people now create 4 billion images per week across ChatGPT, Codex, and the API (announcement; The Verge notes that is up from the 3 billion weekly figure OpenAI cited in September).

For everyday users, the practical split is:

Job	Use
Reaction loop, meme, mascot wink, product “glow”	Ask ChatGPT for a GIF
Print-ready poster, logo draft, photo edit with no motion	Still image (same Images tools as before)
Editable vector logo for a website	Still mostly an API SVG job (see appendix)

If you already use ChatGPT for coding agents or models like GPT-6.1 Terra in Work and Codex, Images 2.6 is a separate visual lane: same account, but you talk to Images / describe a GIF rather than picking Terra. Image generation still counts against whatever image and plan limits your tier already has; OpenAI’s Help Center for Images 2.5 said “Existing image-generation limits are unchanged” for that earlier update (release notes, Sept 8, 2026), and the 2.6 announcement does not publish a new ChatGPT per-plan GIF quota. Treat heavy GIF batches the same way you treat heavy still generations: if the app slows you down or shows a limit message, wait or upgrade.

Who this is for (and who can skip)

Do this today if you:

• Make social posts, Slack/Discord reactions, or short product teasers and hate stitching frames in another app.
• Already create images in ChatGPT and want motion without learning video tools.
• Need a looping storyboard for a character or brand mascot before you commission real video (OpenAI’s own use cases).

You can skip (for now) if you:

• Need long video, dialogue, or camera moves — GIFs here are short loops, not a Sora-style movie.
• Only need print stills; Images 2.5-style still generation already covers you.
• Rely only on the Help Center article and refuse to try a feature until that page is rewritten — the GIF UI is live per OpenAI’s announcement, but the Help Center has not caught up yet.

Make your first GIF in ChatGPT (no code)

1. Open ChatGPT where Images works

Images (and now GIFs) are available on web, iOS, and Android, and you can also generate and edit images in Codex (Images in ChatGPT; announcement). Sign in on chatgpt.com or open the ChatGPT / Codex apps.

You do not need a special “GIF mode” toggle in the announcement. OpenAI’s framing is: ask for a GIF the same way you ask for an image.

2. Ask for a GIF in plain language

Type a short request that includes the word GIF (or “animated GIF” / “looping animation”). OpenAI’s public demo prompt was:

can you make a gif of a cat dancing?

(The Verge, describing OpenAI’s example.)

Better prompts for real work add subject, motion, loop intent, and style. Examples you can paste:

Reaction GIF

Make a looping GIF of a friendly cartoon office worker giving a double thumbs-up, front-facing, soft pastel office background, seamless loop, no text.

Product loop

Make a short looping GIF of this product slowly rotating a quarter turn on a clean white table, soft studio light, no logos other than what’s on the product, seamless loop.

Mascot for social

Create a looping GIF of our brand mascot (a small blue bird) waving once, then returning to the start pose. Keep the same character proportions across frames. No captions.

OpenAI’s announcement also shows long “sprite sheet” style prompts used by partners for six-frame character loops (announcement). You usually do not need that complexity in ChatGPT chat: start with a short motion description, then refine.

3. Or upload a photo and ask to animate it

The Verge reports that ChatGPT can create GIFs from a prompt or a photo you upload (The Verge). On web or mobile:

1. Attach a photo the same way you already attach images for edits (Help Center: use + / Add photos & files, or drag onto the composer — see also ChatGPT Image Inputs FAQ for supported upload types).
2. Ask something like: “Turn this into a looping GIF where I nod once and smile, keep my face recognizable, seamless loop.”
3. Wait for generation. Complex requests can take a while; the Help Center notes image generation “may take a few minutes” and that you can keep using ChatGPT while it runs (Images in ChatGPT).

Privacy tip: if you upload a selfie or anything identifying, remember personal ChatGPT accounts may use chats to improve models unless you change training settings. See our walkthrough on stopping ChatGPT and Codex training on your data. Business and Enterprise workspaces default differently; check your workspace policy.

4. Adjust playback speed and frames

After the GIF appears, OpenAI’s demo UI lets you:

• Change playback speed
• View / edit the GIF frame by frame
• Share the result

(The Verge.)

Use that before you regenerate everything:

• Too frantic → slow the playback.
• One weird middle frame → open frame view and edit or regenerate that frame via chat (“fix frame 4 so the hand stays on the desk”).
• Almost right but wrong background → keep the motion, ask to change only the setting.

This is the main value-add beyond OpenAI’s announcement: treat the first GIF as a draft, then use speed and frames instead of burning a fresh full generation every time.

5. Save, copy, or share like any other image

The Help Center’s manage-images steps still apply to what ChatGPT stores under Images:

• Open Images in the sidebar to browse what you made.
• Open an item and use Copy, Save (download), or Share.
• To remove something from My images, delete the conversation where it was created (Images in ChatGPT).

On mobile you can also share a prompt template link so someone else can recreate a still from your prompt (Images in ChatGPT). For GIFs, prefer downloading the finished file when you need the exact loop.

Prompt recipes that work better for loops

OpenAI’s announcement stresses identity consistency (“keeping characters, products, and brand details consistent as you edit”) (announcement). Loops fail most often when the model invents a new face every frame. Steal these constraints:

1. Say “seamless loop” and “return to the starting pose.” That matches how OpenAI describes GIF use cases and how the API guide describes describing “desired loop” in the prompt (API guide).
2. Lock camera and outfit. “Fixed camera, same clothes, same background.”
3. Name one motion. Nod, wave, bounce, rotate — not five actions at once.
4. Ban text unless you need it. Text-in-image is still a weak spot for image models generally (API limitations); GIFs make bad text worse because it flickers.
5. Ask for short. ChatGPT GIFs are for reactions and teasers. If you need a 30-second spot, use a video tool.

OpenAI’s partner examples in the announcement use six-frame sprite-sheet prompts with identical panel sizes (announcement). In ChatGPT you can approximate that by saying “short loop, about half a dozen frames, smooth in-between poses” without pasting a full sprite-sheet brief — then tighten with frame edits.

Still images vs GIFs vs “thinking” images

Keep these Help Center distinctions in mind so you do not chase the wrong control:

• ChatGPT Images (create/edit) is on all tiers (Images in ChatGPT).
• Images with thinking (extra reasoning for harder image jobs) is on Plus, Pro, and Business, with Enterprise/Edu “coming soon” at the time of the Help Center page (Images in ChatGPT).
• Templates and mobile Sketch (@ Sketch) remain useful for stills; Templates are “not yet available in Work mode” (Images in ChatGPT).
• GIF is the new motion path from Images 2.6 (announcement).

If ChatGPT returns a still when you wanted motion, reply: “Animate that as a looping GIF” or start a fresh message that includes the word GIF. If it returns a GIF when you wanted print quality, ask for a still PNG/JPEG instead and use the aspect-ratio picker for the final crop (Images in ChatGPT).

For other “create files inside the chat” workflows on a different product, see our Gemini guide on creating and editing docs, slides, sheets and PDFs inside Gemini — different vendor, similar “finish the asset without leaving chat” idea.

Safety, refusals, and what not to animate

OpenAI’s system card for ChatGPT Images 2.6 says the safety stack builds on Images 2.5 with additional safeguards for GIF creation, because motion can introduce risks that are not obvious in a single still (for example harmful meaning across frames). GIF-specific rules can block some categories more broadly than stills. OpenAI also reports that motion-related violations remain hard to catch perfectly (system card).

Practical rules of thumb:

• Do not ask for sexualized motion, violence, or anything involving minors.
• Expect stricter refusals on GIFs than on similar stills.
• If a request is blocked, rewrite toward a mundane subject rather than trying to “jailbreak” the filter.
• Report policy-violating outputs using OpenAI’s content reporting guidelines linked from the Help Center FAQ (Images in ChatGPT).

Teen and parental-control notes in the Help Center still apply to image create/edit for linked teen accounts (Images in ChatGPT).

Troubleshooting

Symptom	Likely cause	What to try
No GIF controls after generate	Rollout lag, or you got a still	Ask explicitly for a “looping GIF”; update the mobile app; try web
Help Center still says 2.5	Docs lag	Trust the Nov 4 announcement for availability
Face changes every frame	Weak identity lock	Add “same character, same face, fixed camera”; edit frames
Flickering text	Text-in-GIF is brittle	Remove captions; add text later in CapCut/Canva
Generation takes forever	Complex prompt	Simplify motion; keep using ChatGPT while it runs (Help Center)
Hit a limit mid-batch	Plan image limits	Pause; see shared ChatGPT/Codex usage changes if you also burn Work/Codex heavily
Upload rejected	Wrong file type / size	Help Center image-input FAQ: PNG/JPEG/non-animated GIF uploads, 20MB (FAQ) — note that uploading a non-animated GIF is not the same as generating an animated GIF

Optional appendix: API Flare, Sunburst, and GIF fields

Skip this section if you only use the ChatGPT app. Developers get GPT-Image-2.6 Flare (fast iteration) and GPT-Image-2.6 Sunburst (higher detail / consistency), plus markup-guided edits, SVG output, and animated GIF output through the Image API (announcement; image generation guide).

Token pricing (standard, per 1M tokens) for both gpt-image-2.6-flare and gpt-image-2.6-sunburst (API pricing):

Modality	Input	Cached input	Output
Text	$5.00	$1.25	—
Image	$8.00	$2.00	$30.00

OpenAI says Flare can generate in as little as six seconds when quality is set to low; Sunburst is aimed at medium/high quality work with stronger edit consistency (announcement). A separate complexity field (low / medium / high, default high if omitted) controls planning effort and has no separate fee; pair it with the existing quality setting (API guide).

Animated GIF via API (Image API generations only — edits do not support GIF output): set output_format: "gif", describe subject/motion/loop in the prompt, then optionally set (API guide):

Field	Allowed	Default
frame_count	2–64	9
frame_duration_ms	20–2000 (rounded to 10 ms)	200
size	auto or WIDTHxHEIGHT (multiples of 16, within model limits)	auto

Decode b64_json and save as .gif. GIF output does not support streaming, partial images, custom output_compression, or prompt_rewriter (API guide).

SVG: set output_format: "svg" on generations (n must be 1); not available on edits (API guide).

Markup edits: send a transparent PNG overlay as markup with Flare or Sunburst on /v1/images/edits (API guide).

What to do next

1. Open ChatGPT and ask for a simple looping GIF (start with OpenAI’s cat-dancing style prompt, then switch to your brand).
2. Use playback speed and frame edits before regenerating.
3. Download keepers from Images, and tighten privacy settings if you upload photos.
4. Only then wire Flare/Sunburst GIF fields into an app if you need batch automation.

Images 2.6 will not replace a video editor, but it removes the awkward “generate six stills and stitch them yourself” step for the short loops most people actually post. That is the whole point of today’s release — and it works from the same ChatGPT account you already have.