> ## Documentation Index > Fetch the complete documentation index at: https://docs.cogniagent.ai/llms.txt > Use this file to discover all available pages before exploring further. # Generating images > Ask your AI coworker for a picture and get a real file back — header images, diagrams, mock-ups and social cards, saved straight into the task. Ask for a picture, get a file you can use. ## What it does Describe the image you want and your coworker makes it. Not a link to somewhere else — an actual file, saved into the task, ready to download or hand to the next step. Because the image lands in the task's files, your coworker can keep working with it: drop it into a document it's writing, attach it to an email, or include it in a report. ## When you'd use it * A header image for the blog post your coworker just drafted * A simple diagram to make a proposal easier to follow * Product mock-ups for a pitch deck * A social card sized for a campaign * Placeholder artwork so a draft doesn't look half-finished * An illustration for an internal announcement ## Turning it on Open your coworker, choose **Capabilities**, and find **Generate images** in the Built-in tools list. It's on by default. > **Generate images** — Create images from text prompts, saved into the task files. The Generate images row on the Capabilities tab, switched on ## What it looks like The image appears inline in the conversation, and the file appears in the **Files** panel beside it. You can open it full size, download it, or ask for changes — "make it wider", "try it on a dark background" — and get a new version. ## Good to know **Describe the job, not just the subject.** "A header image for a blog post about warehouse automation, wide, muted colours, no text" gets you something usable. "A warehouse" gets you a warehouse. Images are saved as files, so they carry through the rest of the task. Ask your coworker to write the post *and* make the header, and it can do both in one go. ## Common questions Yes. Say what you want different and your coworker generates a new version. The earlier ones stay in the files, so you can go back to one. Sometimes, and not always accurately. For anything where the wording has to be exactly right — a price, a name, a headline — generate the image without text and add the words yourself. This capability creates new images from a description. For working with an image you already have — resizing, converting, combining several into a PDF — that's [Working with code and files](/cowork/capabilities/code-and-files). It's a reasonable one to switch off for a coworker that never needs it — a support agent, a data analyst. Fewer capabilities means a more focused coworker. ## Next steps Resize, convert and combine what you already have. Open an image without downloading it. Where generated images live. Back to the full list.