> For the complete documentation index, see [llms.txt](https://oodlz-ai-studio.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://oodlz-ai-studio.gitbook.io/docs/outbound-agents/awareness/2-articles/image-generation.md).

# Image Generation

3–4 unique in-blog images per article — evocative, varied by angle and composition, with full editorial metadata generated before each image is rendered.

> **3–4 evocative in-blog images per article — each with a different angle, composition, and editorial purpose.**

Generic AI image tools produce one image per prompt with no concept of editorial flow. The Oodlz Image Generation Agent reads the full research brief and produces 3–4 in-blog images per article, each calibrated for a specific position in the post — hero, mid-content, pre-conclusion. Variety is enforced by design: no two images share the same angle, photographic genre, or composition.

## How It Works

The agent reads the article brief, then plans the image set as a whole — not one image at a time. Each concept is assigned a position in the article (hero, mid-content, pre-conclusion), a photographic style, a shot type, and a depth of field. A diversity check ensures the set spans different angles, genres, and compositions; if two concepts collide, one is regenerated.

For every concept, the agent writes the editorial metadata before any image is rendered: placement instruction (which paragraph the image follows), caption, keyword-rich alt text, and a one-line content purpose. The metadata persists with the image throughout the pipeline — the writer references it during article generation, the link pipeline avoids placing links in image captions, and Gutenberg formatting positions the image at the correct paragraph break.

Images render via FLUX 1.1 Pro with professional camera specifications — Hasselblad, Canon R5, Sony A7RV — for photorealistic output. Images are evocative rather than literal: a finance article gets a parent at a kitchen table, not a stock chart of a balance sheet.

## What You Get

* 3–4 unique in-blog images per article — purpose-built, never recycled stock.
* Variety enforced — different angles, depths, and compositions across the set.
* Full editorial metadata per image — placement, caption, alt text, purpose — written before rendering.
* Photorealistic output — professional camera specifications baked into the generation prompt.
* Evocative rather than literal — images carry mood that matches the writing, not data they would have to invent.