PlugWiki

Best AI for image generation, and can Claude generate images?

Claude cannot generate images. Gemini and ChatGPT can, and you can connect an image generator to Claude so it makes pictures for you.

Picks
6
ranked

No. Claude does not make photos or illustrations, though it can draw charts, diagrams and simple SVG graphics with code. For images from a sentence, use Gemini or ChatGPT, which both have image models built in. If you want to stay in Claude, connect an image generator such as Comfy Cloud, fal or Replicate.

Good for

  • People who want a picture, poster or illustration from a sentence
  • Claude users who want images without switching apps
  • Developers who need icons or assets generated from Claude Code
  • Marketers making social images and ad variations

The short answer

Use Gemini or ChatGPT to generate images: Gemini's Nano Banana 2 is on every plan, including free. If you work in Claude, connect Comfy Cloud, fal or Replicate so Claude can call an image model for you.

Our pick: Gemini

Free image limits on Gemini and ChatGPT are not fixed numbers and change with demand, and the generators you connect to Claude bill your own account with them.

Best for

  • Images from a sentence on a free plan: Gemini
  • Images inside Claude Code: fal or Comfy Cloud
  • Icons and simple graphics you can edit as code: Claude's SVG output

Skip it if

  • You expect Claude alone to make a photo or illustration
  • You need images of real people, which vendors restrict

Generate an image in Gemini

Gemini

  1. 1

    Open Images

    Go to gemini.google.com, open the sidebar and click Images. You need to be 13 or older.

  2. 2

    Describe the picture

    Start with a word like draw, generate or create, then say the subject, style, light and shape.

    Prompt
    Generate a photorealistic image of a mountain lake at golden hour, wide shot, mist on the water, 16:9.
  3. 3

    Change it by asking

    Ask for one change at a time in the same chat. On a Google AI plan you can redo an image with Nano Banana Pro.

    Prompt
    Same scene, but add a small red canoe near the shore and make the sky clearer.
  4. 4

    Download it

    Download the image. It is 1K without a Google AI plan and 2K with one.

    Tip: Your limit refreshes every 5 hours until you hit a weekly cap, and images use more of it than text.

Generate images from Claude Code with fal

Claude

  1. 1

    Add fal's server

    In a terminal run claude mcp add --transport http --scope user fal https://mcp.fal.ai/mcp-relay. Use --scope project instead to add it to one project only.

    Uses Claude Code

  2. 2

    Sign in

    Start Claude Code, type /mcp, pick fal and choose Authenticate. A browser window opens to sign in to your fal account.

  3. 3

    Ask for the image and where to save it

    Say the size and the folder so the file lands in your project.

    Prompt
    Make a flat app icon of a paper plane on a blue gradient, 1024x1024, and save it as a PNG in ./assets.

Connect Replicate to the Claude app

Claude

  1. 1

    Get a Replicate API token

    Create a Replicate account and copy an API token. Replicate charges for each model run.

  2. 2

    On claude.ai, add it as a connector

    Open Customize, then Connectors, click + Add, then Add custom connector, and paste Replicate's hosted server address https://mcp.replicate.com. Free plans can add one custom connector.

  3. 3

    Or in Claude Desktop, edit the config

    Open Settings, then Developer, then Edit Config, and add {"mcpServers":{"replicate":{"command":"npx","args":["-y","replicate-mcp"],"env":{"REPLICATE_API_TOKEN":"your-token-here"}}}}. Restart Claude Desktop and check that Replicate shows under the search and tools icon.

  4. 4

    Ask Claude to pick a model and generate

    Claude can search Replicate's image models and run one for you.

    Prompt
    Use Replicate to find a good text-to-image model, then generate a watercolor illustration of a lighthouse at dawn.

Get an icon or diagram straight from Claude

Claude

  1. 1

    Ask for an SVG or a diagram

    Claude cannot paint a picture, but it can draw shapes in code. That works well for icons, logos to iterate on, charts and diagrams.

    Prompt
    Draw a simple flat icon of a paper plane as an SVG, two colors, on a transparent background.
  2. 2

    Save it

    For a chart or diagram Claude draws in the chat, use Copy as image, download it as SVG or HTML, or Save as artifact. This works on every plan on the web and desktop apps, not the phone apps.

Best plugs for this

Ranked by who makes it, then GitHub stars and weekly downloads from our data. fal's and Replicate's own hosted servers are covered in the steps above.

  1. 1Comfy MCP logo
    Comfy MCPOfficialHosted or LocalOAuth

    Comfy Org's own server. Connect Claude to hosted Comfy Cloud with a sign-in, or drive ComfyUI on your own computer.

    317 GitHub stars
  2. 2Hugging Face MCP Server logo
    Hugging Face MCP ServerOfficialLocal

    Hugging Face's own server. Claude can make images through image apps hosted on Hugging Face Spaces.

    925/wk downloads a week
  3. 3nanobanana-mcp-server logo
    nanobanana-mcp-serverLocal

    Uses Google's Gemini image models, Nano Banana 2 by default, from Claude Desktop or Claude Code. Needs your own Gemini API key.

    402 GitHub stars
  4. 4
    Canva MCPOfficialHostedOAuth

    Canva's own connector. Claude can create, fill in and export Canva designs, but it does not make AI photos.

  5. 5fal-mcp-server logo
    fal-mcp-server

    A community server for fal's image, video and audio models. Needs your fal API key.

    61 GitHub stars
  6. 6OpenAI Image logo
    OpenAI ImageLocal

    A community server for OpenAI's image models. Needs your OpenAI API key.

Gemini

Plan: Free includes Nano Banana 2. A Google AI plan adds 2K downloads and Nano Banana Pro for redos.

What it can do

  • Generate images from a sentence with Nano Banana 2
  • Redo an image with Nano Banana Pro on a paid plan
  • Edit the images it makes, for users 18 and over

Limits

  • You must be 13 or older to generate images
  • No published image count; limits refresh every 5 hours up to a weekly cap
  • Free downloads are 1K

ChatGPT

What it can do

  • Generate images with GPT Image 2, sold as ChatGPT Images 2.0
  • Use a Thinking variant for more careful images

Limits

  • OpenAI does not publish a fixed free image count
  • Images carry C2PA data and an invisible watermark
  • Sexual, political or sensitive images of real people are not allowed

Claude

Plan: Any plan. Free can add one custom connector, which is enough for one image generator.

What it can do

  • Draw charts, diagrams and SVG graphics in the chat
  • Call an image generator you connect, such as Comfy Cloud or Replicate
  • Generate image files from Claude Code with fal or Comfy

Limits

  • No built-in image model
  • Inline visuals do not show in the phone apps
  • Connected generators bill your own account with them

Plugs to turn on

Microsoft Copilot

Plan: An eligible Microsoft 365 subscription. Personal, Family and Premium plans have usage limits.

What it can do

  • Create an image from **Create**, then **Create an image**
  • Pick a style, color and size: square, tall or wide

Limits

  • Sizes top out at 1536 by 1024 pixels
  • Microsoft does not name the model or publish limits

Comparison

Use Gemini when you want images on a free plan, or ChatGPT if you already pay for it. Claude only makes images through a generator you connect.

OptionMakes imagesFree useDownload sizeAI label
GeminiThis pageYesYes, limited1K, 2K paidSynthID + visible
ChatGPTYesNot publishedNot documentedC2PA + invisible
Microsoft 365 CopilotYesMicrosoft 365Up to 1536 pxNot documented
ClaudeNo, SVG onlyYesNot applicableC2PA

Questions

Can Claude generate images?

No. Anthropic says Claude does not generate photos or illustrations the way image tools do. It can draw charts, diagrams and SVG graphics with code, and it can call an image generator you connect.

Does Claude Code generate images?

Not by itself, and a request to add it is still open. Connect an image server such as fal or Comfy Cloud, or ask Claude Code for SVG files you can turn into PNGs.

What is the best AI for image generation?

For most people, Gemini or ChatGPT. Gemini's Nano Banana 2 is on every plan, including free, and ChatGPT uses OpenAI's GPT Image 2 model. Inside Claude, connect Comfy Cloud, fal or Replicate.

How do I connect an image generator to Claude?

On claude.ai open Customize, then Connectors, click + Add, then Add custom connector, and paste the server address, such as https://cloud.comfy.org/mcp. Free plans can add one custom connector.

Do the Canva or Adobe connectors let Claude make images?

Not AI images. Canva's connector makes Canva designs, and Adobe's connector edits photos but cannot generate new ones, as an Adobe community manager confirmed.

Can ChatGPT generate images for free?

ChatGPT makes images with GPT Image 2, but OpenAI does not publish a fixed number of free images. Limits change with demand, so expect fewer on the free plan.

Is Gemini's Nano Banana free?

Yes. Nano Banana 2 is on every Gemini plan, with a limit that refreshes every 5 hours up to a weekly cap. A Google AI plan adds 2K downloads and Nano Banana Pro for redos.

Can Microsoft Copilot generate images?

Yes, in the Microsoft 365 Copilot app. Open Create, then Create an image, and pick a style, color and one of three sizes up to 1536 pixels. It needs an eligible Microsoft 365 subscription.

Do AI-generated images carry a watermark?

Usually. OpenAI adds C2PA data and an invisible watermark, Gemini adds SynthID and a visible mark, and Claude adds a C2PA Content Credential to PNG, JPG and SVG files it makes.

Can I make images of real people?

Only within limits. OpenAI says sexual, political or sensitive images of real people break its rules, and Google asks you not to violate other people's privacy or copyright.

Sources

Docs, issue trackers and discussions this page is built from.

Show all 21
  1. Can Claude produce images? · support.claude.com
  2. Custom visuals in chat and Cowork · support.claude.com
  3. Generate and edit images with Gemini Apps · support.google.com
  4. GPT Image 2 · developers.openai.com
  5. comfy-mcp · github.com
  6. Set up fal MCP in Claude Code · fal.ai
  7. MCP server · replicate.com
  8. Gemini Apps limits and upgrades · support.google.com
  9. ChatGPT Images 2.0 image provenance · deploymentsafety.openai.com
  10. [FEATURE] Render and generate images inside Claude Code · github.com
  11. Get started with custom connectors using remote MCP · support.claude.com
  12. Claude builds visuals · claude.com
  13. Hacker News comment · news.ycombinator.com
  14. Hacker News comment · news.ycombinator.com
  15. Hugging Face MCP Server · huggingface.co
  16. nanobanana-mcp-server · github.com
  17. Canva connector · claude.com
  18. Image generation · gemini.google
  19. Create AI-generated images with the Microsoft 365 Copilot app · support.microsoft.com
  20. Check files made with Claude · claude.com
  21. Can not generate images via Claude · community.adobe.com

Updated Nov 1, 2026.