How to Make a GIF in ChatGPT: Images 2.6 Steps, Prompts and Fixes

Jump to a section

How to make a GIF in ChatGPT Images 2.6

OpenAI introduced Images 2.6 on November 4, 2026. The release adds GIF creation to ChatGPT and the API, while the ChatGPT Help Center documents the consumer steps. You do not need to know how to animate individual frames or write code to try the ChatGPT workflow.

1. Open ChatGPT: Use ChatGPT on the web or the iOS or Android app. Open a new chat, or go to the Images area if it is available in your interface.
2. Describe one clear action: Begin with “Make a looping GIF of…” and name the subject and movement. A single readable action—such as waving, blinking, bobbing or wagging a tail—is a good first test.
3. Set the scene: Mention the background, camera position, visual style and what the final moment should do. For a loop, ask the movement to return naturally to its opening position.
4. Or upload a still image: Add a photo, illustration or other still image, then explain which part should move and which details should remain unchanged.
5. Review the result: Check the start-to-end transition, subject consistency and unwanted movement. Ask ChatGPT to revise the animation in the conversation if needed.
6. Adjust and save: Open the GIF to change playback speed, save it to your device or share it.

What to expect from the first attempt

Start with a short, low-complexity action. A character waving once, a candle flame flickering, a paper boat bobbing or a pet blinking is easier to judge than a long scene with several moving objects. If the first result looks wrong, make one change at a time: reduce the motion, simplify the background, or clarify where the action should finish.

Copy-ready ChatGPT GIF prompts

These examples are written for the regular ChatGPT interface. They are prompts, not guaranteed output specifications; the generated motion, duration and frame count can vary. Replace the bracketed details with your own subject.

How to make a ChatGPT GIF loop smoothly

A seamless loop is not simply a sequence that ends. The final pose needs to connect visually to the first one. OpenAI’s launch post includes an example prompt for a six-frame sprite sheet arranged in three columns and two rows. It asks for identical panel dimensions, a fixed camera and a final frame that transitions towards the first. That is a useful prompt design pattern, not a required format for every ChatGPT GIF.

Five prompt details that help

• One main action: Ask for one movement before adding secondary effects.
• Fixed camera: Ask the camera not to pan, zoom or change angle unless that movement is intentional.
• Stable scene: Keep clothing, lighting, background and object placement consistent.
• Natural return: Describe how the action returns to its opening pose, rather than only saying “loop forever.”
• Simple background: Fewer moving details make visual defects easier to notice.

Interactive GIF prompt builder

Enter a subject and action to build a prompt tailored to your starting point. Select ChatGPT for a natural-language prompt, or API planning to include the separate developer settings. The tool runs locally in your browser; it does not send your inputs to OpenAI.

Images 2.6 API: frame count, duration and native GIF output

This section is for developers using the OpenAI Image API. It is separate from the no-code ChatGPT workflow above. OpenAI’s API guide documents native GIF generation through POST /v1/images/generations or POST /v1/images/edits. To create a GIF, set output_format to gif and describe the subject, motion and intended loop in the prompt.

API request example

The following is a request-body example for an application making a server-side API call. It illustrates the relevant settings; it is not JavaScript to paste into a WordPress page. Keep API keys on your server and never expose them in browser-side code.

{
  "model": "gpt-image-2.6-sunburst",
  "prompt": "Create a seamless looping GIF of a paper boat bobbing gently on a pond. Keep the camera fixed and return the boat to its opening position.",
  "output_format": "gif",
  "frame_count": 9,
  "frame_duration_ms": 200,
  "size": "auto"
}

For a photo-led animation, send the still image to the image-edit endpoint and describe what should move. The API returns the animation in the response’s b64_json field; decode that value and write the bytes to a file ending in .gif.

What about API pricing?

OpenAI’s current Image 2.6 API guide lists image output at US$30 per million billable image-output tokens. That is a token rate, not a flat price per GIF. Actual output cost depends on how many billable image tokens the generated frames consume; text and image input can add cost. Check the live OpenAI image-generation guide and cost calculator before estimating a production workflow. Do not multiply the single-frame estimate by a frame count unless the actual usage accounting for your request supports that calculation.

Flare vs. Sunburst: which one should a developer choose?

GIF creation is missing: what should you check?

Do not assume that your account is ineligible just because you cannot see a dedicated GIF control. OpenAI’s launch announcement says Images 2.6 is available globally across subscription tiers, while the Help Center separately says GIF creation is rolling out. That distinction explains why availability can look different between accounts.

1. Try the web version: Open ChatGPT on the web and start a fresh conversation.
2. Update the mobile app: Check the App Store or Google Play for an available ChatGPT update, then reopen the app.
3. Ask directly: Use a plain request such as “Make a looping GIF of a paper boat bobbing on water.” This matches the style of example in OpenAI’s Help Center.
4. Try a still image: Upload a normal still image and ask ChatGPT to animate one clearly identified part of it.
5. Check again later: GIF creation is rolling out. OpenAI has not published a reliable date or manual switch that guarantees activation for a particular account.

Saved reference photos: keep a character easier to reuse

Images 2.6 also introduces saved reference images in ChatGPT. OpenAI’s Help Center describes uploading a photo, asking ChatGPT to save it as a reference, choosing a name, and calling it later with an @ mention. For example, after saving a dog photo as Pepper, a later prompt can refer to @Pepper.

This can reduce repeated uploads when you want to reuse a subject in later image conversations. It is not a guarantee that every generated frame will be pixel-identical, and it is not a reason to save another person’s identifiable photo without permission. Use reference images thoughtfully, especially for customer, colleague or client material.

Why a GIF prompt can be treated differently from a still image

OpenAI’s Images 2.6 System Card explains that a sequence can communicate something that is not obvious from any single frame. Its safeguards therefore consider motion across frames, including movement that could make otherwise ordinary still images convey harmful or sexualized meaning. A prompt may be blocked even when one frame looks harmless. Keep requests respectful, lawful and consistent with OpenAI’s image policies.

The same System Card describes provenance measures that include C2PA metadata and SynthID watermarking across ChatGPT, Codex and the API. These are provenance signals, not a guarantee that every third-party platform will preserve metadata after editing, conversion or re-uploading.

ChatGPT GIF creation FAQ