AI blog image generator: A hero image for every article, in your style.

One hero, plus up to three in the body. The body count comes from length: nothing under 900 words, one image to 1,399, two to 1,999, three above that. Then the topic is checked: legal, financial, procedural and comparison topics get none, because a stock scene adds nothing to a tax explainer, and borderline topics get one fewer than their ceiling.

Yes. Set Hero images to I supply my own in the Brand Hub, or choose Upload now on a single brief, and no hero is generated: add yours on the brief or with Upload your own in the editor. You can still generate one for any article. Upload the heroes you already have as template references, and every generated hero is made to match them. Upload a photograph through the Image Studio and it lands in the site's library, insertable in any article on the site. JPG, PNG, WebP or GIF, up to 8 MB.

Brand Hub, Visual Guidelines. Pick a hero template, and its reference images, plus any rules you add, steer every hero. Leave it on Auto and the prompt is built from your industry, audience, region and brand colours, so a Toronto insurer and a Denver syndicator do not get the same picture. Body images have their own setting: nine looks from photographic to line art, plus an accent colour and any notes you add. When your hero default is a template, Verand looks at its references once and, if they're in a different style from your body images, offers to match them; nothing changes unless you accept.

The one you pick for that template. In the template's second step, six engines from Google, OpenAI, ByteDance and Black Forest Labs each draw it once on a real title from your site, all at the same time, in about 2 minutes, so you can see which one matches your references. A live feed shows each sample as it lands, and it keeps running if you close the window or refresh: the template card shows the comparison running, and asks you to pick an engine when it's done. Picking one makes that template your hero default. Template heroes never invent figures: no made-up percentages, prices or statistics in the image unless they're in the article title. The compliance gate reads the article's text, and this keeps unchecked numbers out of the picture too.

A 1536 by 864 JPEG at quality 85, cropped to 16:9 from whatever the generator returned, with a descriptive filename and alt text on the tag. Same treatment for the hero and for body images, so nothing arrives at a different size or quality than the rest.

No, and that is in the prompt rather than left to chance. Every image request excludes identifiable faces, readable text, signage and business logos, uniforms and insignia, named landmarks and recognisable addresses, and vehicle plates. People appear only at a distance.

The images go with it. On a static site they are committed in the same commit as the page, and if a file cannot be loaded the publish is refused rather than shipping a page with a broken hero. On WordPress they are uploaded to the media library with their alt text, the first one becomes the featured image, and a re-publish reuses what is already there.