Image Generation
Your agent makes the image, not just the words. Generation lives in the editor, so the picture is built from the piece your agent is already writing and lands inline where it belongs — no hunting through a separate tool to find where these images even get made.
How it works
insert_imagegenerates an image from a prompt and inserts it at the cursor- The agent builds the prompt from the surrounding writing — the angle and claim of the piece, not a literal read of the title
- Output is a real image in the document, stored alongside the rest of your content
What you can make
| Vector | Notes |
|---|---|
| Blog feature images | Written to your blog’s image conventions |
| X thread images | One image per body tweet — progression or supporting visuals |
| X article covers | One-image covers sized for long-form X articles |
| OG / social cards | Open Graph cards at the right dimensions |
| Illustrations + icons | Inline supporting art and small icon work |
Consistent characters
For multi-image work — a comic strip, a series — reusable reference images keep a cast visually consistent across every panel. Define the references once; every generation pulls from them.
Model + access
- Runs on Nano Banana (Google Gemini image model). More models are planned.
- Free when you bring your own API key.
- Or generate through the Publish plugin if you’d rather not wire up a key.
Related
- X-Writer uses this pipeline for article covers, thread images, and comic strips
- GitHub Blog Sync commits generated feature images alongside the post
- X API posts the finished piece, images and all