Can Claude generate images? Yes, with a connector (2026 guide)

Short answer: Claude can’t make photos or illustrations on its own. It can read images, describe them and draw simple graphics in code (SVG diagrams, charts, icons). But add an image connector and Claude can call real image models, such as GPT Image or Google’s Nano Banana, and show the finished picture right in the chat.

This guide covers what Claude can do out of the box, how image generation through a connector works, the steps to set it up, and prompts that get good results.

What Claude can do with images by default

• Understand images: read screenshots, charts, handwriting and photos you attach.
• Draw with code: SVG logos, diagrams, flowcharts and data charts it writes as code and renders.
• Write prompts: detailed briefs you could paste into an image tool.

What it doesn’t do on its own is create a photo-real product shot, a poster with a person in it, or an illustration in a painted style. That needs an image model.

How image generation works through a connector

Claude supports the Model Context Protocol (MCP): an open standard that lets it call outside tools. An image connector is an MCP server that gives Claude tools like generate_image, remove_background and upscale_image. When you ask for a picture, Claude:

1. turns your request into a precise brief (subject, light, style, text, shape),
2. picks a model that fits (text-heavy posters need a different model than photo edits),
3. tells you the price before anything that costs a lot,
4. calls the model, waits a few seconds, and shows the image in the chat.

Because the request still goes through Claude, you get its strengths too: it asks clarifying questions, keeps track of your brand and earlier versions, and can edit the result step by step.

Set it up in two minutes

1. Create a free Typeshoot account (100 free credits, no card).
2. Copy your connector URL from the dashboard.
3. In Claude, open Settings → Connectors → Add custom connector, paste the URL and click Connect.
4. Approve access on the Typeshoot page that opens. That’s it: start a new chat and ask for an image.

Prompts that work

Product shot

PromptMake a premium product photo of a matte black water bottle on wet slate, morning side light, shallow depth of field, 4:5 for Instagram.

Poster with text

PromptDesign a poster for a coffee shop called Slow Morning. Headline “SLOW MORNING” in a tall serif, a steaming cup below, cream and espresso colours. Use GPT Image so the text is crisp.

Edit your own photo

PromptHere’s my product photo [attach it]. Remove the background, then put it on a sunny kitchen counter.

Which model does what

You don’t have to choose: say what you want, and Claude picks a sensible default and tells you the cost.

Can Claude edit images?

Yes, through the connector. Attach a photo and ask for the change: swap the background, put a product into a scene, change colours, turn a selfie into a cartoon, or add your logo. Claude uploads the file (uploads are deleted after use) and the result lands in your library.

Tips for better results

• Say the shape: 1:1 for feeds, 4:5 for Instagram, 9:16 for Stories and TikTok, 16:9 for YouTube.
• Put exact on-image text in quotes.
• Ask for 2–4 variations of a hero image, then edit the best one.
• Save your brand once (logo, colours, tone) with a brand kit and say “on brand”.

Next: can Claude make videos?