How to Make Images in Claude

To make images in Claude, you need to connect Claude to an external image tool. Anthropic's own help center says Claude does not generate photos or illustrations the way image-generation tools do. Claude can analyze uploaded images and create custom diagrams or charts, but if you want actual raster images inside the same conversation, the fastest path is to connect ToolRouter and let Claude call the right image workflow for you. That gives you general image generation, product photography, background removal, and image optimization through one connector instead of five separate apps. If you want the broader connector setup first, start with How to Add Connectors to Claude and our roundup of the best MCP tools for Claude.

People keep searching this because Claude is close enough to a general-purpose creative assistant that it feels like it should already do it. Anthropic has also added custom visuals in chat, which makes the line even blurrier. But there is a real distinction: native Claude visuals are HTML and SVG-based diagrams, charts, and interactive elements. They are not the same thing as asking for a finished JPG or PNG hero image, product shot, social graphic, or ad creative.

According to Everypixel, people are creating an average of 34 million AI-generated images per day. According to Grand View Research, the AI image generator market is expected to reach $1.08 billion by 2030, growing at a 17.7% CAGR from 2024 to 2030. The demand is obvious. The question is not whether people want AI images. It is whether they can get them without leaving the Claude workflow they already prefer.

What Claude Can and Cannot Do Natively#

Anthropic's current documentation is pretty clear on the split:

• In Can Claude produce images?, Anthropic says Claude does not generate photos or illustrations the way image-generation tools do.
• In Custom visuals in chat, Anthropic says Claude can create diagrams, charts, and interactive visuals built in HTML.
• In Anthropic's vision docs, Claude is documented as an image input model: it can analyze images you upload, but that page does not position Claude as a native text-to-image product.

That gives you a useful rule of thumb:

If all you need is a quick visual explanation of an idea, native Claude is often enough. If you need a real asset you can publish, ship to a client, upload to Shopify, or drop into a blog post, you need tool access. That is the gap this query is really about.

How to Make Actual Images in Claude#

The no-code version is straightforward.

1. In Claude, go to Settings → Connectors → Add custom connector.
2. Enter:

• Name: ToolRouter
• URL: https://api.toolrouter.com/mcp

1. Start chatting normally.

That works in Claude chat, Claude Desktop, and Cowork. If you are a developer and you prefer the terminal, Claude Code users can run claude mcp add toolrouter -- npx -y toolrouter@latest mcp, but the connector UI should be your default if you just want image workflows working fast.

Once ToolRouter is connected, you do not have to memorize tool names. You can simply ask for the outcome:

• "Create three blog hero image concepts for an article about how to make images in Claude."
• "Turn this rough product photo into a clean studio shot and a warm lifestyle shot."
• "Remove the background from this sneaker photo, then make a 1200-pixel-wide WebP for the website."
• "Make a simple diagram of our onboarding flow and export it as SVG."

Claude figures out which tool to call, and when the task is multi-step, it can chain them in one conversation. That is the part most workaround articles miss. The win is not only "Claude can now generate an image." The win is that Claude can generate, edit, clean up, resize, and package the asset without you context-switching across multiple products.

Which ToolRouter Image Tool to Use#

You will get better results if you think in workflows instead of in one giant "image generation" bucket.

Inside Generate Image, ToolRouter's current catalog gives Claude access to 20+ image models with different strengths. The local tool manifest currently defaults to nano-banana-2, and the catalog also includes newer Google image options such as Nano Banana Pro and Imagen 4 alongside other models optimized for speed, realism, or typography. The practical point is that you do not need to leave Claude every time one model is better at text, product mockups, or faster drafts. Claude can stay in charge of the workflow.

Three Image Workflows That Are Better Than "Just Generate an Image"#

The most valuable use cases are usually chained workflows, not single prompts.

1. Blog Hero Image Workflow#

Use this when you are publishing articles, landing pages, or social cards and need a unique visual that matches the topic.

Workflow: Generate Image → Image Ops

Ask Claude for the concept first, then the file prep:

Generate three warm editorial hero image concepts for a blog post titled "How to Make Images in Claude." Use a physical metaphor instead of a laptop screenshot. After I pick one, export the final as an optimized WebP and JPEG for the site.

That lets Claude create the image, then prepare a production-ready asset instead of leaving you with a raw file that still needs resizing and compression. If you publish frequently, this is a better pattern than bouncing between Claude, a separate image app, and then another optimization tool.

2. Ecommerce Listing Workflow#

Use this when you start with a rough photo and need a listing-ready asset set.

Workflow: Background Removal → Product Studio → Image Ops

Example:

Remove the background from this candle photo, then create one clean studio shot on white and one lifestyle shot on a wooden bedside table with morning light. After that, export both at 1600px wide in WebP.

This matters because product images are revenue-critical. According to Etsy, 90% of shoppers said photo quality was "extremely important" or "very important" to a purchase decision. If your goal is to sell a product, a specialized chain beats generic one-shot image generation almost every time.

3. Social Campaign Workflow#

Use this when you need a whole visual system, not a single image.

Workflow: Faceless Carousels → Generate Image → Image Ops

Example:

Create a 6-slide Instagram carousel about why Claude cannot natively generate photos, but can create images through connected tools. Keep the tone practical, not hypey. Render the slides, then create a matching cover image and export the set in web-ready sizes.

This is where "one connector" really matters. Claude can research the topic, write the slide copy, generate the visuals, and format the output in one thread. You are not managing a writing app, a design app, and a separate image converter just to publish one post.

Best Prompts to Use in Claude#

If you want working prompt patterns instead of generic advice, start here.

General image generation#

Create 3 image concepts for a blog post titled "How to Make Images in Claude."
The article explains that Claude cannot natively generate photos or illustrations,
but can do it through connected tools. Use a warm editorial photography style and
a physical metaphor instead of showing code or app screens.

Product photography#

Take this rough phone photo of my product and create:
1. a clean studio shot on white,
2. a lifestyle shot on a kitchen counter with warm morning light,
3. a square social version with more negative space.
Keep the product itself unchanged.

Background removal plus final export#

Remove the background from this shoe photo, add a subtle realistic shadow,
then export it as a transparent PNG and a 1400px-wide WebP for the site.

Diagrams instead of photos#

Make a clean onboarding flow diagram for our SaaS product.
Use an exportable SVG, not a photorealistic image.

That last example is important because many people searching this topic actually want "something visual," not necessarily "a photorealistic image." In that case, native Claude visuals or Diagram Generator are the better answer than a text-to-image model.

Why One Connector Beats Piecing Together Separate Image Apps#

You can absolutely cobble together a Claude image workflow manually. People do it every day. They ask Claude for a prompt, paste that prompt into a separate image generator, download the result, use another app to remove the background, and then export it somewhere else for web.

That works. It is also slow.

Here is the real comparison:

The practical benefit is not only convenience. It is context retention. Claude remembers what the image is for, what style you want, what the previous version looked like, and what the next step should be. That matters a lot when you are iterating on a hero image, adjusting a product shot, or producing a whole asset set that needs to stay consistent.

If your goal is simply "make Claude spit out an image," almost any workaround can do that. If your goal is "stay in Claude and finish the job," the workflow matters more than the first render.