ChatGPT Images
ChatGPT Images is OpenAI's image generation and editing, living inside ChatGPT rather than as a standalone app: describe an image in a conversation, or upload one and describe the change, and the result comes back in the same thread. OpenAI deprecated the DALL·E models and removed them from its API on 12 May 2026, and ChatGPT Images is the successor a buyer actually uses today.
The current generation is available on every ChatGPT tier, with an editor for selecting and revising regions, aspect-ratio control, and a library that keeps everything you have made. Edits follow instructions conversationally, each request refining the last, and generation continues in the background while you keep chatting.
For developers the same capability ships as the GPT Image model family in the API, priced by usage. It is an ecosystem capability rather than a design platform: quick, conversational and everywhere ChatGPT is, with production design workflow deliberately left to the design tools.
- Cost
- Freemium (Free tier + paid plans)
- Ease
- Model
- Hosted service
- Checked
- August 2026
Prices, plans and model versions change fast: this is a mid-2026 snapshot; check the tool's official site for the latest.
Best for
- Creating and editing images inside a ChatGPT conversation
- Instruction-following edits that refine turn by turn
- Region-level edits with the built-in selection editor
- Image generation included on every ChatGPT tier
- Builders reaching the same capability through the GPT Image API family
Less suited to
It is a capability of ChatGPT, not a design platform: brand kits, templates, asset libraries and team workflow live in the design tools, and production creative still graduates to them. Anyone searching for DALL·E should know the models are gone from OpenAI's API, removed on 12 May 2026, and this is what replaced it.
The most art-directed, cinematic image quality remains the aesthetic benchmark's ground, and lettering-heavy design work has its own specialist; this optimises for accessibility and instruction-following over either extreme.
Costs & data, in short
Not sold separately: image generation is a capability of ChatGPT's tiers, available on all of them, so there is no second subscription to buy.
Images are generated inside ChatGPT, so prompts and outputs follow the data controls of the ChatGPT plan in use.
The current generation is available on every ChatGPT tier, including the free one.
In practice
How ChatGPT Images is used, area by area.
Jobs
Design & creative
For most people the fastest route to a usable image is the assistant they already have open. ChatGPT Images generates and edits in the flow of a conversation: describe what you want, look at what comes back, and say what to change, with the editor there when a specific region needs work. The craft skills the dedicated tools assume, prompting conventions, layers, masks, are replaced by plain instructions, which is exactly the trade a non-designer wants. The ceiling is real but so is the reach: it is on every ChatGPT tier, on web and mobile, with no new tool to adopt.
Example tasks
- Generate an image mid-conversation and refine it by replying
- Edit an uploaded photograph by describing the change
- Fix one region with the selection editor
- Re-crop to a different aspect ratio without regenerating
- Keep everything generated in one searchable library
Limits
Production creative on a brand system, or the single most striking frame, belong to the dedicated design and image tools.
Compares
| vs | Pick ChatGPT Images when | Pick the other when |
|---|---|---|
| Nano BananaFull comparison → | ChatGPT Images is the same ambient-assistant idea for people who already live in ChatGPT | the ecosystem you already use is Google's rather than OpenAI's |
| IdeogramFull comparison → | ChatGPT Images fixes one region with the selection editor rather than regenerating around it | logo directions should be explored before anybody commits to vector work |
| MidjourneyFull comparison → | plain instructions replace the prompting conventions, layers and masks the dedicated tools assume, which is exactly the trade a non-designer wants | mood and style control are what the image is being chosen for |
Tasks
Image editing
Editing is where instruction-following pays most
Editing is where instruction-following pays most. Upload an image, say what should change, and ChatGPT Images applies it, with a selection tool for region-level work, aspect-ratio regeneration, and undo when a request lands wrong. Each turn refines the last, so an edit session reads like a conversation rather than a tool tutorial. The boundary is craft: compositing, colour pipelines and professional retouching stay with the photo editors, and the selection tool's own documentation is honest that highlights are not always precise and edits may extend beyond the selected area.
Example tasks
- Describe an edit to an uploaded image in plain language
- Select a region and revise just that area
- Remove or replace elements conversationally
- Regenerate at a new aspect ratio
- Iterate an edit until it reads right
Limits
Precise professional retouching and compositing belong to the dedicated photo editors.
Compares
| vs | Pick ChatGPT Images when | Pick the other when |
|---|---|---|
| Photoroom | ChatGPT Images covers general conversational editing across whatever the chat throws up | the job is listing-ready e-commerce imagery at volume |
| Nano Banana (Gemini image)Full comparison → | ChatGPT Images is for the edit that can be described in a sentence when perfection is not the bar, with a selection tool for region-level work and regeneration at a new aspect ratio | the change must land without collateral damage to the rest of the scene |
Image generation
ChatGPT Images earns its place on this hub by being where the most people already are. Generation happens in conversation on every ChatGPT tier, and the succession matters for anyone arriving here on an old name: OpenAI removed the DALL·E models from its API on 12 May 2026, and the GPT Image family now does the work, with ChatGPT Images as the consumer surface. Against the dedicated generators the trade is reach against ceiling: no subscription beyond ChatGPT itself, no new interface, and background generation while you keep working, at the cost of the art-directed extremes the benchmark models chase.
Example tasks
- Produce a concept image while thinking through an idea
- Generate variations by conversational refinement
- Illustrate a document or post without leaving the assistant
- Try several aspect ratios of one concept
- Move to the GPT Image API when volume outgrows chat
Limits
A defined visual style across a campaign, or frontier aesthetic quality, want the dedicated generators.
Compares
| vs | Pick ChatGPT Images when | Pick the other when |
|---|---|---|
| MidjourneyFull comparison → | ChatGPT Images wins on reach and convenience inside the assistant most people already have | the frame itself is the deliverable and its aesthetic ceiling is the point |
| Nano Banana (Gemini image)Full comparison → | ChatGPT Images generates in conversation on every ChatGPT tier, which is why it is where most people already are, though a defined visual style across a campaign wants the dedicated generators | control matters more than spectacle |
| IdeogramFull comparison → | ChatGPT Images produces a concept image while you are still thinking an idea through, which is a different moment from producing a finished one | the free tier should be used to learn the model before anything is paid for |
Where to start
Not sure what to adopt first?
Five quick questions about your job, task and constraints. We'll suggest your top three tools, plus the one to try first.
Alternatives
Same category, different strengths.
Common questions
Is ChatGPT Images free?
There's a free tier to start; paid plans add capacity and features.
Where does ChatGPT Images fit best?
ChatGPT Images fits best in Design & creative and Image generation; see its practice notes for how.
Before sharing confidential or personal data, check this tool's data-governance and training policies. They differ between providers and can change.
Last checked: August 2026