Skip to main content
Image Generation creates illustrations from an article’s image plan and places successful results into its draft. Review the images alongside the text before publishing. If an image is missing or unsuitable, choose the action that fits the problem: retry a failed image, create a new concept, edit an existing image or upload your own.

How images are planned

The writer is asked to plan images before writing the article. Each planned image has a purpose, such as a hero or section illustration, a description and a detailed rendering prompt. The planning instructions recommend two to four images; they do not enforce that number in every finished post. Your project’s image style and saved brand colors can guide those prompts. They do not confirm that the generated scene, typography or business details are accurate. The description is an input to the process, rather than a visual inspection of the finished image. Alt text is derived from the rendering prompt and a compatible plan description. Read it against the actual image, especially after changing the concept. A description of the intended scene can still be wrong for what the generator produced.

Aspect ratios

An image plan can carry an aspect-ratio label. The current image service maps the following labels to requested pixel sizes. These are request settings, not a shape picker in the blog editor or a guarantee about the returned file.

Aspect labels and current requested image sizes

Plan labelRequested pixels
16:9, also the default1536 × 1024
4:31024 × 768
1:11024 × 1024
3:4768 × 1024
The default’s 1536 × 1024 request is not an exact 16:9 canvas, despite the stored label. Check the actual image and how your website crops it. Creating a new concept uses the service’s default size; editing an existing image asks the service to choose a size from the reference. Neither action promises an unchanged crop.

How images appear in your post

The article text is saved with placeholders such as IMAGE_1 before the image phase starts. The blog can then show Text ready, images pending while image generation continues. The normal image phase runs up to two image requests at a time and attempts to replace each successful placeholder with its image URL. The featured-image choice prefers a successful image marked as the hero. If none has succeeded, it uses the first successful image. A section illustration can therefore occupy the featured slot while the intended hero is missing; check the featured image rather than assuming the first completed image is the hero. The normal image phase sends a completed draft to Generated, pending review, including when some or all planned images failed. Missing images remain a separate review task. Other failures, cancellation or interrupted work can produce a different status, so read the displayed progress and error before taking another action.

Image styles

Choose Image Style in the Images tab of Preferences. The saved style guides the prompts used for blog images. Review the resulting look; a style thumbnail is an example of the choice, not a preview of your article’s image. Brand & Text intentionally allows typography. Other generation prompts receive guidance against invented names, logos and readable signage. Those instructions do not establish that an image contains no text or that all branding is correct. Check signs, packaging, screens and clothing in the finished scene. For your own logo, use Business Info > Sources > Brand images > Brand logo on images. Upload the logo, enable the toggle and choose its position. This is separate from the image-style preference. The pipeline attempts to overlay an enabled, saved logo after generation; if that step fails, it can retain the image without the logo. Review the actual placement and presence.

When an image doesn’t come out

A failed image can leave an IMAGE_N placeholder and a missing-image notice. The text may already be saved and reviewable. Check whether the notice concerns missing images, the article’s generation or publishing, because those need different recovery actions. Use Retry images when the writing is useful and the recorded image attempts failed. If the message reports an invalid prompt or a content-safety rejection, inspect that reason before repeating the same request. A retry can fail again, and an original plan entry must still be available for the retry to run. See Blog Generation for the full article workflow and Regenerating Content for the retry’s status restrictions.

Retrying failed images

Click Retry images on the blog’s missing-image notice. This action reads the recorded failed entries and reuses their original image-plan prompts and aspect labels. It skips the article-writing step and attempts to place successful results into the saved content. Retrying is additional image-generation work, not a free recovery promise. The image pipeline attempts to record usage after the provider responds, before local decoding and storage. A later storage problem does not mean no image request ran. No fixed retry price is established by this guide; check your account’s usage and the displayed billing errors. A retry does not adopt a new prompt entered for a separate image edit. If the original prompt keeps failing, use a suitable replacement workflow rather than expecting another retry to use different instructions. Inspect the post again after the result appears, including its featured image and remaining placeholders.

Regenerating a single image

Open the blog, choose Edit Article and select the image. Under What should AI do?, choose one of these modes:
  • New concept creates a fresh image. Enter your idea in New image concept, adjust Modern to Realistic if useful, then click Generate new image. With no new idea, the request uses available original planning context, alt text, the title or keyword as context.
  • Edit current sends the existing image as a reference. Describe the change in Edit instructions, then click Apply edit. An edit needs a nonempty instruction and an existing image belonging to the blog.
The style slider guides the request toward a modern, abstract or photographic treatment. Composition preservation is an instruction for Edit current, not a guarantee about the returned result. A custom prompt alone does not choose the mode in this editor. Keep either prompt within 1,000 characters, which the API accepts. The request starts background work; a starting response is not the finished image. Wait for the created/updated result or error, then check the draft. Successful work updates the selected image in the saved blog and also updates the featured/header URL when it matched that image. Copy anything you need to retain before replacing it.

Using your own real images

For a photo or screenshot that should replace a selected image, choose Replace Image in Edit Article. The editor accepts JPEG, PNG, WebP or GIF and shows a 10 MB limit. It uploads the file and saves the replacement into that selected slot; it does not automatically replace the first image in every article. To add an image below text, select one text section and use Add image below this section. This adds the image after the selected section. Review the alt text as well as the picture: a replacement retains the old image’s alt attribute, and this add-image path starts with an empty one. An editor upload stores a blog image directly. It is different from adding a reusable item under Sources > Brand images, where you can review a description and its Active setting. Active, described, blog-enabled library images can be offered to the writer as candidates; adding one does not force its placement in a future post. For connected-tool upload interfaces, see theStacc MCP and the upload_image reference. Check that interface’s placement and storage contract instead of assuming it behaves like Replace Image.

Delivery, format, and safety

The generated-image pipeline attempts WebP conversion before storing a public image URL. It can fall back to PNG if conversion is unavailable or fails. This does not establish a fixed size saving, a loading-time result or an improvement in search rankings for your page. The generation request uses provider moderation, and a rejected request can produce a content-safety message. Moderation is not verification of the depicted facts, permission to publish or professional approval of medical or legal content. Review the image’s provenance and intended use before publishing it. A successful draft replacement is also separate from an update on your live website. Check the saved article and featured image, then follow your publishing workflow and inspect the live result.