Media enrichment

The Media enrichment tab adds pictures and videos to every post: original images created by an AI generator, free stock photos from Pixabay, Creative Commons images from Google, and YouTube videos. All of them can be combined.

WordPress Dashboard › RSS Retriever

Click to enlarge

AI image generation

Select one Generator: OpenRouter, OpenAI, Flux, Midjourney or Stable Diffusion, or None. Each generator needs the matching API key on the Accounts page. Its settings appear below the selector.

Image position

Above content / Below content

One image at the top or at the bottom of the post.

Store in "thumb" custom field

The image is not inserted into the text; its URL is saved in the thumb custom field. Set Post thumbnails to Create from "thumb" custom field to use it as the featured image.

Batch insert

Inserts several images into the article: one before each occurrence of the tag in Insert images before tag (for example h2), up to Number of images. With %gpt_image_prompt% in the prompt, a separate prompt is written for every section from the text that follows the image, and earlier prompts are taken into account so the pictures do not repeat themselves. The best way to illustrate long AI articles.

Replace first image

Replaces the first image of the post with a generated one. If the post has no image, the new image goes above the content.

Replace all images

Replaces every image of the post with generated images.

Writing image prompts

Describe the image in plain language. Placeholders and Spintax work in every prompt. Three approaches:

• A fixed style with the title: A watercolor illustration for an article titled "%post_title%".
• Let AI read the article: %gpt_image_prompt%. The Default AI model writes a prompt about a concrete scene of the text.
• Your own analysis instruction: %gpt_prompt[Describe one photo-realistic scene that illustrates this article, max 30 words: %post_content_notags[300]%]%.

The Image class field of every generator sets the CSS class of the inserted <img> (default aligncenter) for alignment, size or effects.

OpenRouter

One key gives you many image models: FLUX, Google Gemini image models, OpenAI image models, Seedream and others.

• AI engine: the OpenRouter model ID, for example black-forest-labs/flux-1.1-pro or google/gemini-2.5-flash-image. Quick Select lists the image models currently offered by OpenRouter. Default: openrouter/auto.
• Image assignment: the prompt.
• Aspect ratio: 1:1, 16:9, 9:16, 3:2, 2:3, 4:3, 3:4, 4:5, 5:4, 21:9 or auto. Default: 16:9.
• Image resolution: Standard (1K), High Res (2K), Premium (4K) or Low / Draft (512), if the model supports it.

OpenAI

• Model: GPT-Image-2 (default), GPT-Image-1, DALL·E 3 or DALL·E 2.
• Image assignment: the prompt.
• Image size: the sizes offered by the selected model. Default: 1792x1024.

OpenAI rejects prompts that break its content policy; such images are not generated.

Flux

FLUX is known for photo-realistic images and low moderation, which makes it a strong choice for news illustrations. It is available through GoAPI and CometAPI.

• API provider: GoAPI (default) or CometAPI.
• Model: the FLUX models offered by the selected provider, such as flux1-dev or flux1-schnell.
• Prompt and Negative prompt (what must not appear).
• Image size in pixels (GoAPI; width × height must not exceed 1,048,576; default 1344 × 768) or Aspect ratio (CometAPI; default 16:9).

Midjourney

Midjourney has no public API; RSS Retriever uses it through GoAPI or CometAPI.

• Prompt.
• API provider: GoAPI (default) or CometAPI.
• Process mode: Relax, Fast (default) or Turbo. Faster modes cost more.
• Aspect ratio, for example 1:1, 4:3 or 16:9 (default).

Stable Diffusion

• Engine: Stable Image Ultra (default, highest quality), Stable Image Core (fast, no prompt engineering needed), Stable Diffusion 3.5, Stable Diffusion 3 or Stable Diffusion SDXL v1.0.
• Image description: the prompt. If it is empty, no image is generated.
• Negative prompt: elements that must not appear (not used by SD3 and SD3.5).
• Style preset (Core and SDXL): 3d-model, analog-film, anime, cinematic, comic-book, digital-art, enhance, fantasy-art, isometric, line-art, low-poly, modeling-compound, neon-punk, origami, photographic (default), pixel-art or tile-texture.
• Aspect ratio (Ultra, Core, SD3, SD3.5) or Image size and Steps (SDXL: 10 to 150 diffusion steps, default 50; more steps, more detail).

Pixabay

Free, full-size images from Pixabay’s library of millions of photos and illustrations. Needs a free Pixabay API key.

• Pixabay image position: None, Above content, Below content, Store in "thumb" custom field or Replace first image.
• Query: a keyword such as coffee cup. Pixabay matches keywords, not sentences, so a full post title rarely finds anything. Separate several keywords with commas to pick one at random; %gpt_keyphrase_en% often works well.
• Query language: the language of your keywords.
• Type and Orientation.
• Min width and Min height in pixels.
• Image number to choose: 1 to 200 from the search results, or 0 for a random one.
• Image class.

Google Images

Images found on Google Images. Only images with a Creative Commons license are searched.

• Google Image Search position: None, Above content, Below content, Store in "thumb" custom field or Replace first image.
• Google search query: default %post_title%. Placeholders such as %categories%, %gpt_keywords_en% or %gpt_keyphrase% are useful here.
• Min width, Min height and Image class.

YouTube video

• YouTube video position: None, Above content, Below content, or Store in "thumb" custom field, which saves the URL of the video’s preview image to use as the featured image.
• YouTube search query: default %post_title%. %gpt_keyphrase% usually finds more relevant videos.