Generate Images in AI Chat: GPT ImageGen and Nano Banana
How to generate images in the same AI chat window using Create Image, which models MultiChats runs (GPT ImageGen and Google Nano Banana), and what the free allowance covers.
You are mid-conversation, working through a product idea, and you want a picture of it, right there in the thread. The fastest way to do that is to generate images in AI chat directly, in the same thread you were already typing in, without tab-hopping to a separate image tool and losing your train of thought. In MultiChats that lives behind a button called Create Image, and it runs five models across two families: GPT ImageGen and Google Nano Banana. Here is how it works and where the edges are.
What it means to generate images in AI chat
Most image tools make you leave. You open a different product, paste a prompt into a different box, wait, download, then carry the file back to wherever you were actually working. That context switch is small but it adds up, and it kills the rhythm of an idea.
Create Image collapses that loop. You hit the button, describe what you want in plain language, and the picture arrives as a message in the same chat. The text you wrote a minute ago is right above it. The follow-up you want to type next is right below. Your prompt history, your earlier notes, the brand color you mentioned three messages up: it is all still there, because you never went anywhere.
If you have only ever used a standalone generator, the difference feels small for the first image and large by the tenth. Iteration is where it pays off. "Make the background warmer." "Same character, now from the side." "Drop the text overlay." You are talking to the picture the way you talk to the model about everything else.
The models: GPT ImageGen and Google Nano Banana
Create Image runs a small lineup of models, and picking the right one matters more than people expect. Here are the two families on offer:
GPT ImageGen: OpenAI's image model, currently in two versions (GPT ImageGen 2 and GPT ImageGen 1.5). It is the one I reach for when a prompt is dense with instructions, when there is text to render inside the image, or when I need it to actually follow a long, fussy description instead of doing its own thing.
Google Nano Banana: Google's image model, and yes, that is the real name. It comes in three variants: Nano Banana, Nano Banana 2, and Nano Banana Pro. It is fast, it is good at clean edits, and it holds a subject steady across changes, which is exactly what you want when you are iterating on the same idea over and over.
You do not have to memorize which is which. The honest workflow is to try one, and if the result is not landing, switch to the other and run the same prompt again. Different models read the same words differently, and the second take is often the keeper. That same mid-conversation flexibility is the whole point of putting all the AI models in one app, and image generation works the same way.
Which one should I use?
If you want | Start with |
|---|---|
Text rendered inside the image | GPT ImageGen |
Strict adherence to a long prompt | GPT ImageGen |
Fast edits, consistent subject | Google Nano Banana |
Lots of quick variations | Google Nano Banana |
The free allowance and what Pro adds
On the web, free users get 2 image generations every rolling 30 days; on mobile, image generation needs a paid plan. Two on the web is enough to kick the tires, see how Create Image feels, and decide whether having it inside your chat is worth keeping. It will not carry a real project, so treat it as a taste test rather than a working tool.
Pro at $20.99 a month (or $199.99 a year) opens up image generation properly, alongside 25+ chat models, web search with citations, voice calls on mobile, memories, and folders to organize your chats. Pro+ at $34.99 a month raises the ceiling rather than adding a separate image feature: 3,000 message credits a month instead of Pro's 1,500, unlimited memories, and a larger file-size limit for video analysis on Gemini models. On either plan, each request produces one image, so a batch of variations means running the prompt a few times with small tweaks.
Worth a quick note on cost, since image models charge differently than chat. A generated image has its own pricing logic on the model side, separate from how words are counted. If the per-request economics of AI ever feel opaque, our explainer on what an AI token is is the clearest place to start, even though images are billed on their own terms.
How Create Image differs from a separate image tool
A dedicated generator like Midjourney or a standalone Stable Diffusion setup will go deeper on raw control: aspect ratio presets, seeds, style references, the full knob panel. If you are producing finished art at volume, those tools earn their place. No argument there.
Create Image is built for a different moment: the one where the image is part of a larger conversation you are already having. A few things follow from that:
Context comes free. The model already read your earlier messages, so you can say "an icon for the app we just discussed" and it knows what you mean.
No tool-hopping. Text answer, image, code block, table: they all live in one thread, in order, ready to scroll back through later.
One subscription, not two. You are not paying for a chat tool and an image tool separately. Image generation rides along with the plan you already have.
The honest tradeoff: you give up some of the deep control a specialist tool offers in exchange for speed and continuity. For ideation, mockups, thumbnails, social posts, and "show me roughly this" moments, that trade is usually worth it. For a final print-resolution campaign asset, reach for the specialist. Knowing which job you are doing is half of choosing well, which is the same logic behind picking the right best AI chatbot for any given task.
A few prompting habits that help
Image prompts reward specificity the same way text prompts do. A short list of things that consistently improve results:
Name the subject, the setting, and the mood in that order, then add style words last.
If you want specific text in the image, put it in quotes so the model knows it is literal.
Iterate in small steps. Change one thing per follow-up so you can see what actually moved the result.
If a model keeps missing the same detail, switch families and rerun the exact prompt. Sometimes the other engine just gets it.
FAQ
Can I generate images in a chat app?
Yes. In MultiChats you generate images in AI chat with the Create Image button, and the picture appears as a message in the same thread you are typing in. There is no separate canvas or second app to open. You can then keep editing it by replying in plain language.
Which image models does MultiChats support?
Five models across two families: GPT ImageGen 2 and GPT ImageGen 1.5 from OpenAI, and Google's Nano Banana, Nano Banana 2, and Nano Banana Pro. GPT ImageGen tends to win on text inside images and strict prompt adherence; Nano Banana is quick and good at consistent edits. You can switch between them and rerun the same prompt to compare.
How many free image generations do I get?
On the web, free users get 2 image generations every 30 days. On mobile, image generation requires a paid plan, so there is no free allowance on phones. To generate regularly you will want Pro at $20.99 a month, which opens up image generation along with the full model lineup. Pro+ at $34.99 a month doubles the monthly message credits, from 1,500 to 3,000, if you generate and chat heavily.
What is Nano Banana?
Nano Banana is Google's image generation model, and the unusual name is genuinely its name. It is known for fast generation and for keeping a subject consistent across edits, which makes it a strong pick when you are iterating on the same image many times rather than starting from scratch each prompt.
The short take
If your images are born inside conversations, idea sketches, mockups, a thumbnail you need right now, then generating them in the same chat window is the move. Create Image keeps GPT ImageGen and Nano Banana one button away, your context stays intact, and it all sits under the subscription you already pay for. Try it free on the web first, then see the full plan breakdown on our pricing page when you are ready to generate at full speed.