How to Generate Images and Videos on Claude (2026 Guide)

Ask Claude to draw something and it will politely decline. Claude has no built in image generator. It does speak MCP though, and that changes the answer: connect Morphix once and Claude generates images and videos inside the chat you already have open.

The plumbing behind this is mature. In December 2025, when Anthropic donated MCP to the Linux Foundation, the protocol had passed 97 million monthly SDK downloads and 10,000 active servers (Anthropic, Donating the Model Context Protocol, 2025). This guide covers the one minute setup and what Claude can do once it's connected.

Can Claude generate images?

Not on its own. Claude is a text model: it can sketch diagrams and SVG artifacts, but it can't produce photos or rendered art. What it does have is first class support for MCP, the Model Context Protocol, an open standard that lets AI agents call outside tools.

That distinction is the whole trick. An image generator exposed as an MCP server becomes a tool Claude uses directly. You describe the shot in plain language, Claude picks a model from the Morphix catalog, writes the prompt, and the finished image lands in the conversation. No tab switching, no uploads, no copying prompts around.

And the protocol is safe to build a workflow on. On December 9, 2025, MCP moved to neutral governance under the Linux Foundation's Agentic AI Foundation, backed by Anthropic, OpenAI, Google, and Microsoft (Linux Foundation, AAIF announcement, 2025).

How do you set it up?

Three steps, about a minute, no code. The fastest path is the one click install, which opens Claude with everything prefilled and skips straight to signing in. Doing it by hand is barely slower:

1. Open Claude's connector settings. That's Settings, then Connectors, then Add custom connector.
2. Name it Morphix and paste the server URL: https://mcp.morphix.pro/mcp
3. Click Add and finish the Morphix sign in. The free plan works, no card needed.

If Morphix doesn't show up in a conversation after adding it, check the connector toggle behind the plus button in the chat input, and confirm the sign in finished. On other clients the setup is one command instead: the MCP page lists it for Claude Code, Cursor, Codex, and Gemini CLI.

A working session, in practice

We use the connection while building Morphix itself, so this workflow is tested daily on our own product. A typical ask from a working session sounds like this: "Generate a wide hero image of a foggy brutalist building, photoreal, 16:9, for the architecture page." Claude calls the server, checks the model catalog, picks one suited to photorealism, and returns the finished image with a link into your library.

Nothing about the request is special syntax. Because Claude reads the same conversation you're working in, it folds in context you would never bother to retype into a web form. That means the page the image is for, the tone of the copy next to it, and the aspect ratio the layout needs. In our experience that context is worth more than any prompt trick.

However, agents aren't magic. For fine control over reference images, seeds, or batch runs, the web tools are still the faster surface. The practical split: quick assets in chat, deliberate art direction in the app.

How do you generate videos on Claude?

The same connection, the same kind of ask. Describe the clip and Claude generates video from text, or hands a starting image to the model and animates it. Video generations take longer than images, so Claude kicks them off and reports back when the result is ready.

Because Claude can also browse your generation library, the two halves chain together. Yesterday's hero image becomes today's opening frame. Ask for an image, refine it in a message or two, then say "animate this" and you have motion.

The models behind the connection

Claude gets the full Morphix catalog, the same one the web app uses. If the names below mean nothing to you, that's fine. Each one is a generation engine from a major AI lab, each with its own strengths, and you never have to memorize them.

For images that means Seedream, ByteDance's photoreal specialist, Flux from Black Forest Labs, a favorite for stylized art, Nano Banana, Google's editing focused image model, and GPT Image from OpenAI. For video it means Sora 2 from OpenAI, Google's Veo 3.1, Kling 3.0 from Kuaishou, and ByteDance's Seedance 2. Browse the complete list on the models page.

Better still, you don't have to choose. Claude can explore the catalog itself and pick a fitting model for each task. For example, ask for a photoreal product shot and you'll get a different model than for a flat illustration. Prefer hands on control? Every knob is exposed in the web app's image tool and upscaler, up to 4K.

Ready to try it? Grab the URL from the MCP page and ask Claude for an image. Your first one usually lands before you've finished writing the next paragraph.

Frequently asked questions