Brooke Wright
Brooke Wright · @wright_mode
FREE WALKTHROUGH

Set up the new ChatGPT image features

My full walkthrough from the reel. Save a photo under an @name, make GIFs, get SVG logos and fix one spot in an image. Every click, plus copy-paste prompts.

5 short lessons27 copy-paste promptsReal screenshotsChecked 09/11/2026
Something went wrong. Please try again.

No spam. Unsubscribe anytime.

Free walkthrough: the new ChatGPT image features Join the membership → Start lesson 1
Wright Mode — Free walkthrough

Set up the new ChatGPT image features

Everything from the reel, step by step. Save yourself and your product under an @name, make GIFs, get SVG logos and fix one spot in an image without starting over. Then put it all together into a brand kit.

5 short lessons27 copy-paste promptsReal screenshotsAll ChatGPT plans

The lessons

OpenAI released Images 2.6 on 04/11/2026. It's in ChatGPT on every plan, worldwide, on your computer and your phone. That includes Australia and the Free plan.

OpenAI release note dated Nov 4, 2026: ChatGPT Images 2.6 and GPT Image 2.6 Sunburst and Flare. It lists GIF creation rolling out in ChatGPT, saved reference images reused with @Pepper, and in the API: SVG generation, animated GIFs and markup-guided editing for GPT Image 2.5 and 2.6 models.
Tap the image to zoom. Screenshot of OpenAI's release notes (openai.com/products/release-notes), entry dated 04/11/2026. Taken 09/11/2026.
The four things from the reel, as at 09/11/2026
FeatureWhere it livesStatus
Saved photos with @nameChatGPTNew on 04/11/2026
GIFsChatGPTRolling out. Some accounts don't have it yet.
SVG logos and iconsOpenAI's API (the developer side). Not a button in ChatGPT.New on 04/11/2026
Markup editingChatGPT already had it (the Select tool). New in the API.In the API from 04/11/2026

The API is the developer side of OpenAI. You pay per image and send requests with a little bit of code instead of typing in a chat box. Lesson 3 shows you how to use it without writing the code yourself.


LESSON 1 OF 5

You upload a photo once and give it a name. In any later chat, you type @ and that name, and ChatGPT uses the photo. No re-uploading. This is the one I'd set up first, because every other lesson gets easier once it's done.

OpenAI help article section, Save and reuse a reference photo: upload a photo and ask ChatGPT to save it as a reference, choose a name, then type @ followed by its saved name in a later chat.
Tap the image to zoom. OpenAI's own instructions, from Images in ChatGPT in the help centre. Taken 09/11/2026.
1

Pick the photo

For you: a recent photo, face clear and facing the camera, good daylight, no sunglasses, just you in it. For a product: the whole product, label readable, on a plain background.

2

Open ChatGPT and start a new chat

Use chatgpt.com on your computer or the ChatGPT app on your phone. Add the photo to the message box the way you normally add a photo.

3

Ask ChatGPT to save it

Paste this with the photo, swapping the name in brackets. Pick a short, one-word name. You'll be typing it a lot.

Save this photo as a reference called [YourName], so I can use it in later chats.

OpenAI's steps from here: choose a name for the reference, then save it. Check it's the name you want before you save.

4

Test it in a fresh chat

Start a new chat and type @ followed by the name you saved, then what you want. OpenAI's example is "Make a watercolor portrait of @Pepper", where Pepper is a dog.

Only save photos you're allowed to use

Yourself, people who've said yes, and products you own or sell. OpenAI's own image guide says the same: when you show a real person, confirm you have permission to use their likeness.

5 prompts for on-brand photos

Each prompt uses @YourName or @YourProduct. Paste the prompt, delete that placeholder, then type @ and your saved name yourself. OpenAI's help page says to type the @, so type that bit rather than pasting it. Anything in [square brackets] is yours to change.

Make a professional headshot of @YourName for my website and LinkedIn. Keep the face, hair, skin tone and features exactly as they are in the reference photo. Don't smooth the skin or change the shape of the face. Outfit: [a navy blazer over a white T-shirt]. Expression: relaxed smile, looking straight at the camera. Soft daylight from a window on the left, plain warm-grey background. Square, head and shoulders, photo-realistic.
Show @YourName working at a timber desk in a bright home office with a few green plants and a laptop. Outfit: [a rust-coloured linen shirt]. Laughing and looking slightly away from the camera, like a candid photo. Keep the face, hair, skin tone and features exactly as they are in the reference photo. Natural morning light. Portrait 4:5 for Instagram.
Show @YourName speaking on a small stage, holding a microphone, mid-sentence, with warm stage lights and a blurred audience in the foreground. Outfit: [an emerald green dress]. Keep the face, hair, skin tone and features exactly as they are in the reference photo. Landscape 16:9. Leave the right third of the image empty so I can add a title later.
Make a styled flat lay of @YourProduct, shot from directly above on natural linen. Around it: [a sprig of eucalyptus, a small ceramic dish] and soft morning shadows. Keep the product's shape, colours, label and every word on it exactly as they are in the reference photo. Don't add any other logos or text. Square, product in the centre, photo-realistic.
Put @YourProduct in a gift-guide photo for an Aussie summer Christmas. It sits on a sunny timber outdoor table, half-wrapped in brown paper with twine and a sprig of native greenery, with a bowl of cherries blurred in the background. No snow. Keep the product's shape, colours, label and every word on it exactly as they are in the reference photo. Portrait 4:5.
Bonus: you and your product in one photo

OpenAI's examples only use one saved name per prompt. If this mixes the two up, upload the product photo in the same message instead of using @YourProduct.

Show @YourName holding @YourProduct at a market stall, smiling at a customer just out of frame. Keep the face, hair and skin tone exactly as in the first reference, and the product's shape, colours and label exactly as in the second. Soft afternoon light. Portrait 4:5.

How consistent is it, really?

OpenAI says Images 2.6 is "better at preserving details across edits". Better isn't the same as perfect. Its developer guide still says the model "may occasionally struggle to maintain visual consistency for recurring characters or brand elements". Check every face and every label before you post. If one's off, the fix is in Get unstuck.


LESSON 2 OF 5

Ask for a looping animation, or upload a still image or a sketch and ask ChatGPT to make it move. OpenAI says GIF creation is rolling out, so it may not be on your account yet. If it isn't, do lessons 3 and 4 and check back.

OpenAI help article section, Create and edit a GIF: ask for a looping GIF, upload a still image to animate, open the GIF to adjust playback speed, save or share it. ChatGPT can't edit an animated GIF you upload.
Tap the image to zoom. From OpenAI's Images in ChatGPT help article. Taken 09/11/2026.
1

Ask for a looping GIF

Say "looping GIF" and describe what moves. OpenAI's example: "Make a looping GIF of a paper boat bobbing on the water." To animate your own logo, product or sketch, add the image to the message first.

2

Open the GIF to set the speed

Click it on your computer, or tap it on your phone. That's where OpenAI puts the playback speed control, plus Save and Share.

3

Change it in the same chat

Describe the edit ("slower", "make the wave bigger", "cream background"). ChatGPT can't edit an animated GIF you upload, so keep working in the chat where it made the GIF.

4

Save it and use it

Save it to your device. Canva takes GIF uploads, so you can drop it straight into a design.

From OpenAI's Images 2.6 announcement: a pen drawing of an owl on lined paper next to the animated cartoon owl made from it, mid-wave.
Tap the image to zoom. This is the "upload a sketch and make it wave" bit from the reel, in OpenAI's own example. Screenshot of Introducing Images 2.6 (04/11/2026), taken 09/11/2026.

4 GIF prompts

Make a looping GIF from my logo (attached). The letters pop in one at a time with a small bounce, then the whole logo holds still for a beat before it starts again. Keep the logo's colours, shapes and spelling exactly the same. Plain white background, square.

For the mascot, draw it on paper, take a photo and add the photo to the message.

This is a sketch of my brand mascot (attached). Turn it into a friendly cartoon character and make a looping GIF of it waving at the camera. Keep the shape and features from my drawing. Soft colours, plain cream background, square.
Make a looping GIF of the product in this photo (attached) turning slowly on a plain white turntable. One full turn, then it loops. Keep the product's shape, colours, label and every word on it exactly the same. Soft studio light, square.

For the reaction GIF, save your mascot first with lesson 1, as YourMascot or whatever you call it.

Make a looping reaction GIF of @YourMascot doing a happy little dance while confetti falls, for when a client books a call. No words or letters anywhere in the image. Plain [mint green] background, square.

LESSON 3 OF 5

An SVG is a logo file made of shapes instead of pixels, so it stays sharp at any size, from a website tab to a banner. Canva calls it "great for web graphics".

The catch

SVG output is in OpenAI's API, not in the ChatGPT app. OpenAI lists it under "More ways to create and edit in the API", and its developer guide puts it on the Image API. There's no SVG button in ChatGPT as at 09/11/2026. So there are two ways to do this lesson.

From OpenAI's developer guide: two SVG logos for a Field and Flour bakery made from the same prompt, one by GPT Image 2.6 Flare and one by GPT Image 2.6 Sunburst.
Tap the image to zoom. Real SVG logos from OpenAI's image generation guide. Both models got the same prompt. Taken 09/11/2026.

Way 1: a clean logo from ChatGPT (no setup)

You get a PNG with a transparent background, not an SVG. For Canva, social posts and slides, a big transparent PNG does the job. You can also start from a template: open the sidebar, select Images, then Templates. OpenAI gives a logo as one example of what a template can make.

Make a logo for my business, [Business name]. We [what you do] for [who you help]. Style: flat, simple shapes. No gradients, shadows or textures, so it still looks clear when it's tiny. Colours: [#hex] and [#hex]. Text: the words “[Business name]” spelled exactly like that, and no other text. Transparent background, square.

When it's done, open the image and select Save. In Canva, open Uploads in the editor and upload the file.

Way 2: a real SVG file, made through the API

This one has a one-time setup, needs at least US$5 of OpenAI credit, and runs in Claude Code (Pro plan or higher) or the Codex app. Never used Claude Code? Lessons 1 to 3 of my free editing course get it open in a folder. You won't write any code. Claude does that part.

Show me the Way 2 setup
1

Add credit to OpenAI's developer side

Sign in at platform.openai.com and open the billing overview. Select Add payment details and choose an amount. The minimum is US$5. Use auto-reload is on by default, so turn it off unless you want it topping up your card. Credits expire after a year.

2

Make an API key

On the API keys page, create a new secret key and copy it. Treat it like a password.

3

Verify if OpenAI asks

OpenAI's guide says you may need to verify your organisation before you can use its image models. If it asks, it may want a government photo ID (passport or driver's licence), a selfie or your business details.

4

Make a Logo Lab folder and open it in Claude Code

Make a folder called Logo Lab in Documents. Open it in Claude Code (choose Local) or in the Codex app, then paste this:

Set up an SVG logo maker in this folder using OpenAI's Image API. Explain each step in plain English. I'm not a developer. 1. Make a file called .env containing the line OPENAI_API_KEY= and open it in TextEdit, so I can paste my key after the equals sign and save it myself. Don't ask me to paste the key into this chat, and never show it back to me. 2. Write a small script called make_svg. It sends my brief to POST https://api.openai.com/v1/images/generations with model gpt-image-2.6-sunburst, output_format svg and n 1, using the key from .env. It decodes data[0].b64_json and saves the result as a .svg file in a folder called Logos, with a short name based on the brief. 3. If you need to install anything, tell me what it is and why, and wait for my yes. 4. Run one test with this brief: a flat orange circle icon. Open the SVG in my browser so I can see it, and tell me the token usage the API reported. 5. If anything fails, show me OpenAI's exact error in plain English and stop.

After this, you only paste the prompts below into the same chat. Each one makes a file in your Logos folder.

4 SVG prompts (Way 2)

Use make_svg to make this SVG: a logo for [Business name], a [what you do] in [your city]. A simple symbol of [a sprouting seed] above the words “[Business name]”, spelled exactly like that. Flat colours only: [#hex] and [#hex]. No gradients, shadows or textures. Transparent background.
Use make_svg to make this SVG: a square monogram icon with the letters “[BW]” for my website tab icon and profile picture. Bold, simple shapes that are still readable at 32 pixels wide. Flat colours only: [#hex] letters on a [#hex] rounded square. No gradients, shadows or textures.
Use make_svg to make 4 matching icons, one SVG at a time: [a calendar], [a speech bubble], [a lightbulb] and [a coin]. Same style on all 4: simple outline icons with the same line thickness and rounded corners, one colour [#hex], transparent background, no text. Save each one with a clear name.
Use make_svg to make 5 matching icons for my Instagram highlight covers, one SVG at a time: [a question mark] for FAQ, [a star] for reviews, [a gift] for offers, [a camera] for behind the scenes and [a heart] for clients. Same style on all 5: chunky outline icons, one colour [#hex], centred with plenty of space around each one, transparent background, no text.

Use your SVG in Canva or on your website

Make a PNG copy of each SVG in my Logos folder, 1200 pixels wide with a transparent background, saved next to the SVG. Tell me first if you need to install anything, and wait for my yes.

LESSON 4 OF 5

You highlight the bit you want changed, then say what to change. Everything else should stay put. OpenAI calls this image markup and says it was "already available in ChatGPT" before this update. What's new on 04/11/2026 is that developers got it in the API too.

From OpenAI's developer guide: a kitchen photo with a red circle drawn around a striped vase on the table, and the edited version where only the vase has changed to blue.
Tap the image to zoom. What markup does: circle one thing, change only that. This example is from OpenAI's image generation guide and was made through the API. In ChatGPT you do the same thing with the Select tool. Taken 09/11/2026.

On your computer

  • Select an image in your chat, or in Images in the sidebar. The editor opens.
  • Select Select and highlight the area you want to change.
  • Undo or Redo to fix your highlight, or Cancel to start over.
  • Describe the change, then select Save when you're finished.

On your phone

  • Tap an image ChatGPT made to open it full-screen.
  • Tap Select and highlight the area. The slider changes the size of the selection tool.
  • Tap Next.
  • Describe the change in the conversation panel.
OpenAI help article section, Use the editor: Select to highlight the area to edit, Aspect ratio, Undo or Redo, Cancel and Save. Highlights are not always precise.
Tap the image to zoom. The editor buttons, from OpenAI's Images in ChatGPT help article. Taken 09/11/2026.

OpenAI's own warning: "Highlights are not always precise, and edits may extend beyond the area you selected." So every edit below says what must stay the same.

3 example edits

Highlight the area first, then paste the matching prompt.

Remove the [cords and the coffee cup] in the area I highlighted. Fill the space with more of the same desk. Change nothing else in the image.
In the area I highlighted, fix the spelling so it reads “[Wright Mode]” exactly. Same font, colour, size and position. Change nothing else.
Change the colour of the [blazer] I highlighted to [deep teal, #396a6b]. Keep the fabric, the fit, the face and everything else exactly the same.
For developers: markup in the API

In the API, markup is a separate transparent PNG with your circles or scribbles on it, the same width and height as the original image. It goes to the image edit endpoint as markup, with supported GPT Image 2.5 and 2.6 models. Don't combine it with a mask. OpenAI charges the same as any other edit, with no extra fee.


LESSON 5 OF 5

This puts lessons 1 to 4 together. About 20 minutes of your time, plus waiting while images generate. OpenAI says an image can take a few minutes, and you can keep chatting while it works. Do it all in one chat, and make a Brand kit folder on your computer for the downloads.

1

Brief ChatGPT on your brand (2 minutes)

Fill in the brackets and paste this first:

I'm making a small brand kit in this chat. Use this brand for every image I ask for here. Business: [name] What I do: [what you do] for [who you help] Colours: [#hex], [#hex], [#hex] Feel: [warm and down-to-earth, a bit cheeky] Never: [stock-photo handshakes, neon, glitter] Don't make an image yet. Reply with a one-line summary of my brand.
2

Save @YourName and @YourProduct (3 minutes)

Lesson 1, steps 1 to 3. Skip any you've already saved.

3

Make your logo (4 minutes)

Use the Way 1 prompt from lesson 3. If a letter comes out wrong, fix it with edit 2 from lesson 4.

4

Two photos of you (4 minutes)

Lesson 1, prompts 1 and 2.

5

Two product photos (4 minutes)

Lesson 1, prompts 4 and 5.

6

One GIF (3 minutes)

Lesson 2, prompt 1: your new logo, looping. If GIFs aren't on your account yet, leave this one for later.

To download each image, open it and select Save. Everything you make is also kept under Images in ChatGPT. Give the files plain names like logo-teal.png and headshot-blazer.png.

Check it before you post


OpenAI says Images 2.6 is "available globally" in ChatGPT across all subscription tiers. Every plan gets image creation. The difference is how much.

ChatGPT plans in Australian dollars, from chatgpt.com/pricing on 09/11/2026
PlanPriceImages, in OpenAI's words
FreeA$0Limited image creation
GoA$13 a monthMore image creation
PlusA$30 a monthMore complex and accurate image creation
ProFrom A$155 a monthUnlimited and faster image creation (subject to abuse guardrails)
From OpenAI's Images 2.6 announcement, Pricing and availability: Images 2.6 is available globally today in ChatGPT and Codex across all subscription tiers on desktop, mobile, and web. GPT-Image-2.6 Sunburst and Flare are available in the API.
Tap the image to zoom. From Introducing Images 2.6, 04/11/2026. Taken 09/11/2026.

My take: start on the plan you've already got. Free is enough to test lessons 1, 2 and 4. If you hit the image limit halfway through your brand kit, that's when Go or Plus earns its keep.


Open the problem that matches yours. Where there's a prompt, paste it into the same chat.

I don't have GIFs yet

OpenAI is rolling GIFs out, so some accounts get them later. Update the ChatGPT app on your phone and try chatgpt.com on your computer too. If you still get a still image back, it hasn't reached your account yet.

Typing @ and my name doesn't bring up my photo

Type the @ and the name yourself instead of pasting them, and check you're using the exact name you saved. Still nothing? Start a new chat, upload the photo again and ask ChatGPT to save it as a reference.

The face doesn't look like me
The face in this image doesn't look like @YourName. Make it again and match the reference photo exactly: face shape, eyes, nose, smile, skin tone and hair. Don't smooth the skin. Keep the outfit and setting you already made.

If that doesn't fix it, save a clearer photo: front-on, good daylight, nothing covering your face.

My product's label is wrong

Highlight the label with the Select tool (lesson 4) and paste this:

Make the label in the area I highlighted match the label in the @YourProduct reference photo exactly: same words, spelling, colours and layout. Change nothing else.
The edit changed more than I highlighted

OpenAI says highlights aren't always precise. Undo, highlight a smaller area, then name what has to stay:

Go back to the version before that edit. Only change [the mug]. Keep [the face, the desk and the window light] exactly as they were.
ChatGPT won't edit the GIF I uploaded

That's expected. OpenAI says ChatGPT can't edit an animated GIF you upload. Upload a still image of it instead, or go back to the chat where ChatGPT made the GIF and ask for the change there.

make_svg gives me an error

An error mentioning credit_balance_exhausted means you're out of credit. Top up in the billing overview. If it mentions verification, that's lesson 3, step 3.

make_svg just failed. Show me OpenAI's exact error and explain it in plain English. If it's about billing, my key or verification, tell me which page to open on platform.openai.com. Try one fix at most, then stop and tell me what to do next.
Canva won't take my SVG
Make a Canva-friendly copy of [file name].svg: under 3 MB, saved as SVG 1.1, flat colours only, no gradients or textures, with shapes of the same colour grouped together. Save it next to the original with -canva on the end of the name.
I want to delete a saved photo

OpenAI's help page doesn't explain this yet (checked 09/11/2026). Photos saved for shopping try-ons live in Settings, then Personalization, then Reference photos, so look there first. If yours isn't there, ask ChatGPT: "How do I delete my saved reference called [name]?"


THE MEMBERSHIP

My membership is where we build this stuff together.

Three live calls a month: one live build with me, and two co-work and Q&A calls. All recorded. Plus 250+ lessons and resources.

Use this walkthrough to make your first brand kit, then bring what you made to a call.

Join the membership →

Already a member? Bring your brand kit to the next co-work call.

Back to lesson 1 →