Claude Can Now Generate Images (And They Look Stunning)
Claude can finally generate images — better than Gemini, with no prompt engineering needed. Here's how Amplifiers made it happen.
You know the one complaint about Claude that never goes away? That it can’t generate images.
That if you want a visual, you have to leave Claude, open Gemini or ChatGPT, generate the thing, come back.
Every AI tool has its gap. For Claude, that was the gap. Until recently.
Claude image generation is finally here, through Amplifiers (the AI blew my mind MCP).
And here’s the part that sounds like a stretch until you see it: the images come out better than what you get using Gemini directly, even though Amplifiers runs on the same model Gemini does, Nano Banana Pro.
If you’ve had the AI blew my mind MCP since launch day, less than three weeks ago, you already know we shipped it with image generation from day one.
But the only way it could show you an image was by dropping a link into the chat. Functional. Not exactly pretty.
So we went back in.
We rebuilt the whole image generation experience, and the result looks so native to Claude you’d forget it’s coming from an external MCP.
No more switching between apps. No more paste-and-pray. Image generation lives right where you already work.
But don’t take my word for it. Let me show you.
Since this article was published, we've also added ChatGPT's GPT Image 2 to Amplifiers, which is now the best image generation model out there, beating Nano Banana 2.
So you can now use both models inside Claude to generate images.
What’s in this article
What's new in Claude image generation (via Amplifiers)
Here’s a quick rundown of everything that shipped in this update:
Images in chat — generated images now show up right inside Claude, no more chasing a link to see what you made
Refining images — tweak the current image with a small instruction (swap the background, change a color, adjust the text) without starting from scratch
Downloading images — save the final image directly from the chat with one click
Uploading reference images — drop your own image into the chat to guide the generation, whether that’s a logo, a product shot, or a style reference
Style galleries — curated visual styles you pick from instead of describing in words. Three galleries live right now (but more galleries are on the way):
YouTube Thumbnail Styles — 12 styles in 16:9, built to read clearly at small grid size in feeds and search results
Infographic Styles — 12 styles for shareable, scannable infographics
Podcast Cover Styles — square cover art that works across platforms
The style galleries and the prompts behind them are what do the work to make your images look polished and professional. To use them, you need to be a paid subscriber to AI blew my mind.
Features are always easier to see than to read about. So let’s generate a few images in Claude with Amplifiers, and you’ll see everything in action.
One thing to keep in mind before we go in: this is beta. We’re just releasing it, so some things might not work as expected, and some of you might run into errors here and there. If that happens, open a new chat and try again or email me at daria@aiblewmymind.com.
We’re going to keep improving it based on your feedback and our own testing, so it gets better and more reliable over time.
How to generate stunning infographics in Claude
I recorded my screen going through the whole process, in case you’d rather watch than read through the steps and screenshots below:
And here’s the step-by-step of everything it does.
I started by asking for an Amplifier to create an infographic for my Thursday article on Claude Code routines, something I could share on LinkedIn or Substack Notes.
It came back with a couple of questions: aspect ratio, and what I wanted in the footer for branding. You can also mention colors here if you want them baked in.
Because I’d already given it the article URL, it pulled the content from there and suggested text for the infographic. I could approve it or keep iterating. If you don’t give it a URL or any context about the topic upfront like I did, it will also ask you about the topic and content.
This part already existed in Amplifiers v1.
Now comes the new stuff: you choose your infographic style from the style gallery.
I picked hand-drawn notebook and waited a bit for it to generate.
Once an image is generated, you get three options right in the chat:
Download — if you’re happy with the image
Try another style — if you want to go back to the gallery and pick a different one
Refine — if you want to tweak the current one
I went with refine. I asked it to swap the cream background for white.
Now let’s see how different styles would handle the same content.
I clicked “Try another style” and generated the infographic in Isometric 3D and Paper Cutout. Same content, completely different looks.
Pretty cool, isn’t it? You can now see images right inside Claude, and use the arrows to scroll through every variation you generate, whether you got there by refining or trying another style.
One note on refining: it works best for small changes. Color tweaks, background swaps, small text edits. If you want something bigger, like adding your brand colors or rewriting most of the text, write that change in the Claude chat directly instead of using the refine box. You’ll hit fewer errors that way.
How to create click-worthy YouTube thumbnails in Claude
Let’s say I’m also doing YouTube content on the same topic (Claude Code Routines) and I want a thumbnail for it.
I stayed in the same chat where I’d just made the infographics, and asked Claude to help me create a YouTube thumbnail for a video on Claude Code routines.
Here’s the video demo:
Same as before, all I did was tell Claude to use an Amplifier to create a YouTube thumbnail.
It asked me a couple of questions for context first, so it doesn’t waste your generations on something that isn’t aligned with what you want.
I told it I wanted a thumbnail without my face on it. It came back with two concept variations, each with a short description of what would be inside (left image below). I picked one, and then it pulled up the thumbnail gallery so I could choose a visual style (right image).
I picked the bold podcast clip style from the gallery. It generated the thumbnail in 16:9, which is the default for YouTube.
Then I told it I actually did want to upload a portrait of myself. A new prompt box appeared where I could pick a file straight from my computer.
This wasn’t possible before. The old AIBMM MCP couldn’t see what you were uploading, so the only workaround was to share a link to your image. Now you just select the file.
Once the upload went through, I kept iterating and created more variations of the thumbnail.
You’ll notice this time I didn’t use refine. I gave feedback directly in the chat instead. Both work. Use whichever feels more natural for the change you’re trying to make.
Everything I showed you also works without the style gallery. If you want a visual that doesn’t have a gallery yet, you still get:
Uploading reference images
Seeing the image directly in the chat
Refining it
Downloading it
Your input turned into a professional image prompt behind the scenes
The style gallery is an extra layer on top. The rest works across the board.
Just keep in mind you'll need a paid subscription to unlock all the premium image generation prompts.
Why Claude image generation beats Gemini (even though it’s still Nano Banana Pro)
Even though Claude’s image generation runs on Nano Banana Pro (Google’s image generation model, same as Gemini), the images you create inside Claude come out better. Three reasons:
Higher resolution. Amplifiers runs through API calls, so the file you download is much higher quality than what Gemini gives you. Try it yourself: generate the same image in both and scale them up. The Gemini version pixelates fast. (This is also why I used to skip Gemini and go to Google AI Studio when I wanted better quality images, before Amplifiers existed.)
No watermark. Gemini adds a watermark called SynthID to every image. Amplifiers gives you the image without it, the same way you would get it if you went to Google AI Studio and used the Nano Banana API directly.
No prompting work, no app switching, fewer steps. Before, if you wanted to turn an article, a topic, or any input into a visual, you had to:
Leave Claude and go to Gemini (or whatever image tool)
Summarize your input into the exact text for the image
Write a detailed prompt for how the visual should look
Hope it came out right (usually hit or miss if you weren’t good at image prompts)
Now Claude handles all of it. Amplifiers includes a library of image generation prompts built for different formats, and each one takes whatever you give it and turns it into an expert-level prompt that produces great images.
You stay in Claude, you skip the prompting work, and there are way fewer steps to think about. Everything is built in.
How to set up Amplifiers
Setup takes just a few minutes, whether this is your first time or you already had it installed. Below, I’ll walk you through both paths, plus a few things some of you have been asking about when it comes to logging in.
If you’re setting it up for the first time
Step 1. Add the Amplifiers connector in Claude
Open Claude in the desktop app. Go to Customize in the left side panel, click Connectors, then the + icon. Select Add custom connector. Name it AI blew my mind and paste this URL: https://mcp.aiblewmymind.com. Click Add.
Step 2. Connect it
Go back to the Connectors section and scroll down to Not Connected. Find the AI blew my mind MCP and click on it. Hit Connect. This takes you to auth.aiblewmymind.com where you’ll create your account. Make sure you use the same email you subscribed to AI blew my mind with. Then click Always allow for all tool permissions.
Step 3. Add your Gemini API key
Your first 3 image generations are on me. After that, you need to add your own API key to keep generating images.
Here’s how to set it up:
Go to Google AI Studio and sign in with your Google account.
Click Get API key in the left menu.
Click Create API key, then Create API key in new project. Google sets everything up behind the scenes in a few seconds.
Copy the key and store it somewhere safe.
Head to auth.aiblewmymind.com and add your API key there.
Done. Your Gemini API key is now connected to Claude through Amplifiers. When you generate images inside the desktop app, you only pay for credits on the images you generate. Nothing else.
Step 4. Add your OpenAI API key (for Images 2)
If you also want to generate images using ChatGPT’s model, you’ll need to add an OpenAI API key. Same process, different platform:
Go to platform.openai.com and sign in with your OpenAI account.
Click API keys in the left menu.
Click Create new secret key. Give it a name if you want (optional).
Copy the key and store it somewhere safe.
Head to auth.aiblewmymind.com and add your OpenAI API key there, alongside the Gemini one.
Done. You now have both image generation models connected. When you want to generate an image, just mention in your prompt whether you want to use Nano Banana or ChatGPT, and Amplifiers handles the rest.
Generate with Nano Banana or ChatGPT — your choice
Now that you have both image generation models inside Claude, you can pick which one to use every time you generate.
The only thing you need to do is mention it in your prompt. Say “generate an image with Nano Banana” or “generate an image with ChatGPT”, and Amplifiers routes to the right model.
This also means you can generate the same image with both models and compare which output you like best, just like I did in the demo below.
If you already had Amplifiers set up
You need to reconnect it to make sure you pick up all the new tools:
Go to Customize → Connectors in Claude.
Find the AI blew my mind connector.
Disconnect it.
Remove it.
Add it again and connect it, same as a new user (steps above).
It can work without doing this, but reconnecting is the safest path to catch all the new tools and avoid running into errors.
A few things to know about the login (lots of you have been asking)
Your Amplifiers login is separate from your Substack login. The two aren’t connected, since Amplifiers is an AI blew my mind product and Substack is a different platform. So you’ll need to create a fresh account on auth.aiblewmymind.com, even if you already subscribe on Substack.
Do not use your AI blew my mind LAB account either. If you already created an account in the LAB (my resource hub), that’s a separate login system from auth.aiblewmymind.com. You’ll need to create a new account. (We’ll merge them in the future, but for now they’re two separate accounts.)
Your Claude or ChatGPT email doesn’t need to match your Amplifiers account. You can have Claude and ChatGPT on one email, and then when Claude sends you to auth.aiblewmymind.com to create your Amplifiers account, you use the email tied to your AI blew my mind subscription, even if it’s a different one.
Use the same email you subscribed to AI blew my mind with. Your account needs to match your AI blew my mind subscription email, especially if you’re a premium subscriber. If you use a different email, you won’t have access to the premium prompts and tools, including most of the image generation ones I showed you today.
If you’re a premium subscriber, you might already have an account. Every premium subscriber automatically gets an account created for them. If you never activated it, don’t try to create a new one. Use the Forgot password flow instead to set a new password and log in.
Oh, and one more surprise: Research tools are in too
Did you think image generation was the only new thing in Amplifiers? Not even close. We also shipped a full suite of research tools you can use directly inside Claude.
Here’s what’s in there:
YouTube — pull the full transcript with timestamps (something even Gemini doesn’t do natively), top comments with engagement data, video details, channel profiles, and even search across YouTube
LinkedIn — pull any person’s profile, company page, or a company’s recent posts
Instagram — get post details, reel transcripts, and browse a user’s reels
Twitter/X — pull profiles and a user’s top tweets with engagement metrics
Reddit — get subreddit details, browse posts, and even pull Reddit ad details
Facebook Ads — search the Meta Ad Library, pull ad details with creative breakdowns, or list all ads from a specific company
Google Search — run Google searches and get structured results back inside Claude
Web Search (Grounded) — search the open web with source-grounded results built for accuracy
Company Registry — look up legal entities across 309 jurisdictions and 244 countries
Image tools — remove backgrounds from any image
I wrote a full deep dive on how to use the research tools with real examples here: AI Research Tools Inside Claude and ChatGPT
For now, just know they’re all in there, waiting for you to play with them.
Your turn
I can’t tell you how excited I am about this. My own image generation workflows are so much better now, and I think this is going to unlock a lot for you too.
The prompts behind them do most of the work, so whatever visual you want to create comes out looking really good, without you having to fight with prompt engineering or bounce between tools.
It all happens inside your tool of choice.
One thing to keep an eye on: now that image generation in Claude looks so native (I told you, it feels like part of Claude, but it’s actually Amplifiers), it’s also very easy to generate a lot of images fast. So don’t go wild, and keep an eye on your usage.
Now go try it. Drop a comment below and let me know what you think after you’ve played with it.
And share this post so everyone knows Claude can now generate images.
If you want to get the most out of Amplifiers, upgrade to paid. You’ll unlock every prompt inside Amplifiers (premium ones included), all the premium resources in the AI blew my mind LAB, every weekly premium article, and many other perks.
I haven't used ChatGPT yet because I started with Claude, which has worked very well for me so far. Based on this thread it seems that ChatGPT is better at image generation. What type of account do I need for ChatGPT? I assume the free account has limitations, and that I would need a Pro account for image generation.
Thanks
Hi! I set up the MCP and tried to use the ChatGPT image integration for the first time but got this... "Contact whoever runs the "AI Blew My Mind" plugin (the tool powering image generation here) and ask them to top up the OpenAI API credits on their end. That's the account that's actually running dry."