Which AI chatbot makes your images right in the chat?

ChatGPT is still the most complete assistant for making images in a conversation. You describe a scene, it renders it in the thread, then you adjust in small steps: "keep the character, change the background", "put the headline at the top in caps". The user feedback we went through agrees on two things. It handles long, detailed prompts better than the others, and it keeps a consistent style across a series more easily. On text inside images, opinions split with Gemini, but ChatGPT is less likely to invent a caption you never asked for.

Your plan matters a lot. OpenAI says Free gets limited image creation, Go gets more, and Plus adds "more complex and accurate image creation". For Pro, OpenAI publishes no clear image cap. In the US, Go is $8 a month, Plus $20 and Pro starts at $100. The full plan breakdown is in our ChatGPT review.

The weak spot is how opaque the cap is. OpenAI doesn't publish an image count, not even for Plus. On the official community forum, Plus subscribers describe being locked out for 60 hours, and one shows a 720-hour wait. Our tip: when you plan a series, nail the composition on one image first, then ask for variations one at a time. Every failed try burns a generation.

Gemini gives you the most images without paying. Google states plainly that the free tier includes image generation and editing. Images run on the Nano Banana models: Nano Banana 2 Lite when you're on Flash-Lite, and Nano Banana Pro once you have a Google AI plan (Google notes it's only in select countries). You'll see the difference when you download: 1K without a plan, 2K with one.

Editing is where it shines. You can edit a generated image, upload your own photo to retouch, or merge several images into one. Press comparisons often put it ahead of ChatGPT on realism and speed. One catch: you need to be 13 to generate and 18 to edit.

It has the lowest paid entry here: Google AI Plus is $4.99 a month, AI Pro $19.99. The cap isn't an image count. It's a usage pool based on how complex your request is, refilled every five hours up to a weekly limit. In our Gemini review, we found one image eats about 1% of that pool.

Copilot has something the others don't: it makes images where you already work. Microsoft says you can generate images right from the Copilot Chat pane in Word and PowerPoint. For a client deck, that's one less round trip between a chatbot and your file.

There's a free version, but Microsoft doesn't put a number on its cap. Consumer plans raise it. Microsoft 365 Personal gives higher usage than free for AI image creation, at $9.99 a month or $99.99 a year. Microsoft 365 Premium moves to extensive usage, at $19.99 a month or $199.99 a year. Family is $12.99 a month, but read on before you buy it for images.

The trap is in the help page: AI benefits are only available to the subscription owner. Other people on a Family plan don't get the image allowance. Another thing that keeps coming up on Microsoft Q&A: the daily cap doesn't always reset after 24 hours, and a moderator admits exact limits aren't published. Our tip: check your usage under Services & subscriptions in your Microsoft account before a big session. If you're leaving it, see Copilot alternatives.

Mistral renamed Le Chat to Vibe in 2026, at the same address, chat.mistral.ai. Image generation stayed the same: you describe the image in plain language (subject, style, aspect ratio) and the assistant creates it with Black Forest Labs models, the team behind Flux. You can then ask for changes in the next message, or upload an image to a new chat and describe the edit.

The free plan gives "basic image generation and editing". Pro, at $14.99 a month before tax, gives up to 40x more image generations than Free. Verified students get Pro for $5.99. Mistral publishes its plans in dollars.

Mistral spells out the limit itself: during editing, fine details like small text or complex patterns can get lost or altered. The cap is daily and depends on your plan. For a poster packed with copy, ChatGPT is safer. Our tip: generate the image without text, then add the headline in your design tool. You skip the edit that mangles letters.

Perplexity doesn't have its own image model: it lets you choose. In settings, you pick GPT Image 1 from OpenAI, Nano Banana from Google or Seedream 4.5 from ByteDance, or let Default decide. Nano Banana Pro is for Max users. Handy if you want to compare two renders without paying for two subscriptions.

The free plan gets limited image generations. Pro, at $17 a month billed annually, unlocks a limited amount of high-quality images plus medium-quality options. Max is $167 a month billed annually, or $200 month to month. Limits refill on a rolling basis, not on a fixed date, and regenerating or editing an image counts toward your cap.

Two limits to know. First, Perplexity keeps images from its Free plan and individual Pro and Max plans for personal, non-commercial use (only Enterprise plans allow commercial use): for a client visual or an ad, use another tool. Second, some Pro subscribers report losing image access, with support pointing them to Max. Our tip: if you pay for Pro for research, use images for rough drafts and make the final version elsewhere.