Can Claude analyze images?

What Claude's vision can do

Claude can analyze an image you send: describe its content, read the text in it (screenshot, photo of a document, sign), interpret a chart or table, understand a diagram or a mockup, and answer precise questions about what it sees. It can also combine image and text in one request, for example commenting on a screenshot of a code error or explaining a figure. That is what makes it an assistant able to work on more than plain text.

Concrete use cases

There are many uses: transcribe the text in a photo, extract figures from a scanned table, understand an error message in a screenshot, describe an image for accessibility, analyze a document mixing text and visuals, or get a first take on an interface mockup. For long documents combining text and illustrations (PDFs, reports), its ability to read the whole thing at once is especially handy, as we detail in our guide to analyzing documents with Claude.

What it doesn't do (and its limits)

Claude's vision is for understanding images, not generating them: creating an illustration is the job of image-generation tools, not Claude. On the analysis side, it can be wrong on fine details, misread very small, blurry or handwritten text, or misjudge exact proportions and positions. As with text, it can also confidently assert a wrong interpretation. So verify critical information extracted from an image, especially numbers.

How to use it

On claude.ai, you simply attach an image to your message (for example a screenshot or a photo) and ask your question. The clearer and more legible the image, the better the analysis. Be precise about what you want: transcribe, summarize, explain, extract a specific value. Developers can also send images via the API. For details and currently supported formats, refer to Anthropic's official documentation.

Can Claude analyze images?

Yes, Claude is multimodal: it can see and analyze images you send, describe their content, read the text in a screenshot or photo, interpret a chart or diagram, and answer questions about what it sees. It does not generate images, though: it understands them, and it can be wrong on fine details, so verify critical information.

Can Claude read text in an image?

Yes, it can read and transcribe text in an image (screenshot, photo of a document, sign). Quality depends on clarity: very small, blurry or handwritten text may be misread.

Can Claude generate images?

No. Claude understands and analyzes images but does not create them. Generating an illustration is the job of image-generation tools, a different kind of model.

How do I send an image to Claude?

On claude.ai, attach the image to your message then ask your question; a clear image gives a better analysis. Developers can also send images via the API.