ChatGPT Image 2 Setup And 6 Working Prompts

ChatGPT Image 2 is live, but the real unlock is a thinking-mode setup most people never flip on. This lesson covers the exact setup plus the 6 copy-paste prompts for ad campaigns, ecommerce model shots, packaging mockups, pro headshots, lead magnet design assets, and event posters.

What You'll Learn

• The thinking-mode setup trick that unlocks the model
• 6 copy-paste prompts you can use right now
• Ad campaigns, model shots, packaging, headshots, design assets, event posters
• Pro tips and gotchas so your first outputs land

The Setup (Don't Skip This)

ChatGPT Image 2 is live inside ChatGPT Plus ($20/month) and Team plans. But the real unlock isn't the model itself, it's the thinking mode toggle most people don't know to flip on.

Step 1: Open ChatGPT

Go to chatgpt.com on the web. Image 2 needs the web app for image attachments and reference photos.

Step 2: Switch the model to thinking

Click the model picker at the top of the chat and select the thinking variant. This tells ChatGPT to research and reason before it renders. Skip this and you get generic outputs. Flip it on and the model thinks through your prompt the same way a real designer would brief themselves before opening Photoshop.

Why thinking mode matters: Image 2 is a thinking model. Generic prompts only work if you let it think. Without the toggle on, you get the old image gen behavior. With it on, you get one-shot pro outputs from a half-sentence prompt.

Step 3: Prompt with reference attachments

Drag in any reference image (a product shot, a headshot, a competitor ad). Then describe what you want in plain English. Image 2 figures out the rest.

Use Case Library: Six Prompts

Use case 1: Full ad campaign from one product image

Attach a clean product photo. Get back a full ad campaign matching any brand style.

Take this product and create a full ad campaign with three variants: a hero shot for paid social, a lifestyle shot for organic, and a problem/solution comparison ad. Match the brand feel of [reference brand or website URL]. Make each look like it came from the same campaign.

Use case 2: Ecommerce model images

Attach your product. Get back lifestyle shots with realistic models using it.

Generate three lifestyle photos of a model using this product in real-world scenarios. Make each look like a real iPhone photo, natural lighting, no studio feel. Vary the model age range and setting (home, outdoors, gym).

Use case 3: Packaging mockups

Attach a flat product render or sketch. Get a photorealistic packaging mockup.

Turn this product render into a photorealistic packaging mockup. Show it on a clean white surface with soft natural light, then a second shot with the box opened showing what's inside, then a third lifestyle shot with the product in use. Match this brand aesthetic: [drop website URL or reference photo].

Use case 4: Pro headshots from selfies

Upload 4 to 8 selfies. Get professional headshots that look like a $500 photoshoot.

Here are 8 reference photos of me. Generate three professional headshots: 1) LinkedIn-style on a neutral background, 2) lifestyle headshot in a modern coffee shop setting, 3) editorial-style podcast guest shot. Keep my face identical across all three.

Use case 5: Lead magnet and design assets

Generate hero images, OG previews, blog featured images, and section dividers all in your brand style.

Generate four design assets for a lead magnet about [topic]: 1) a hero cover image with bold text overlay, 2) a section divider illustration, 3) an OG image for the landing page (1200x630), 4) a thumbnail for social sharing. Match this brand style: [drop website URL].

Use case 6: Event posters and print

Generate event posters, flyers, signage, and printables in any style.

Design a poster for [event name] on [date]. Include the title, date, location, and a striking visual. Make it look like it came from a top design agency. Style: [minimalist / editorial / bold typography / etc]. Output at print quality, 11x17.

Bonus: Get playful with it

One test worth trying: a blueprint-style technical drawing of a pet, just to see how Image 2 handles weird requests. The output came out perfect on the first try. The lesson: when you have a model that thinks, you can ask it for almost anything and get something usable on the first attempt.

Pro Tips And Gotchas

• Always toggle thinking mode on. The default model is the old image gen.
• Use clean, high-res references. Grainy inputs produce grainy outputs.
• For brand-matched outputs, drop in your website URL. Image 2 will fetch your colors and fonts.
• If text inside the image comes out garbled, regenerate with "render the exact text: [WORDS]" in the prompt.
• Generate 2 to 3 variants per prompt and pick the best. The model is non-deterministic.
• For headshots and lifestyle shots, ask for "real iPhone photo" to avoid the plastic stock-photo look.
• Image 2 can search the internet mid-prompt. Use it for infographics and topical visuals.

Go Deeper

• Prefer a printable version? Download the PDF: https://drive.google.com/file/d/179tC-qxIktU4WhtJIJQlz9P9U2_V6Zi_/view?usp=drivesdk

Want the whole system installed live, with Cooper on the call? Build Your Own AI, the two week sprint runs the first two weeks of August.

Or get a new working tool every Monday inside The AI Operator community on Whop.