How to use shortcodes in image generation task? | CyberSEO Pro | Support Forum

CyberSEO said
The Media enrichment tab is not designed for prompt generation. Could you provide a specific example of what you're trying to accomplish?  

I wanted to generate a prompt for DALL-E via gpt-4o-mini based on the article content and title with my custom instruction for each rss. If this is somehow possible, I would be grateful, since without this option all the images come out the same

CyberSEO said
In this case you can use a %gpt_prompt[prompt]% placeholder there. It will be replaced with the result of your prompt processed by the AI model set in Default AI model, allowing this result to be used inside other placeholders with multiple nesting or within shortcodes (without nesting).

Alternatively select "Batch insert" in the image generation model position drop-down menu and use it with the %gpt_image_prompt% placeholder as suggested here: https://www.cyberseo.net/blog/new-features-inspired-by-gpt-4o-mini/

Placeholder documentation: https://www.cyberseo.net/content-syndicator/#placeholders

Login to see the quote

  

How to insert title and post body to %gpt_promt[prompt]% tag? 
Like a %gpt_promt[Getenare DALLE Prompt to topic %post_title%]% ? 

CyberSEO said
Technically - yes. But I think your prompt should be improved, because what do you think the AI text model will generate when it sees a prompt like "Genenare DALLE Prompt to topic US and China to roll back tariffs in a major breakthrough".

In your prompt, you should specify exactly what you want. I'm talking about prompt engineering. So your placeholder format is correct for CyberSEO Pro, but the prompt itself will be confusing for the AI model. That's why I suggest you to use the %gpt_image_prompt% placeholder. It already contains a good prompt for DALL-E, Midjourney or any other image generation model.

When %gpt_image_prompt% is used in a single image mode (add image, replace image, etc.), the AI text model analyzes your entire article (not just the title) and generates a prompt asking the image model to create an image of something really relevant to the text.

When used in batch insert mode, the %gpt_image_prompt% will contain a very specific prompt for each text block over which the generated image will be placed, so that all images in your article will be relevant to each particular text they related to.

All of these methods work the same way for both imported third-party feeds and brand-new AI-generated articles.  

Yes, I understand all this, but I still want to generate my own prompt, since I played with it for a long time yesterday, and made the correct request in chatgpt for different news sources to generate a prompt for DALL-E 3, and I have already checked it by API. I just want to have calls to the AI ​​assistant from this field

CyberSEO said
You should use the %gpt_prompt[prompt]% placeholder in this case.  

[12-05-25 02:44:24] Generating dall-e-3 image for "%gpt_prompt[generate dalle3 prompt to generate image for topic: Weekly Security Update: May 4 – May 10 . Use funny or modern IT style depending on the topic]%"

it is ok? It didn't show generated prompt for dalle-3 in logs.

CyberSEO said
No, that's not okay. There is an error in the placeholder name - please check it carefully.  

Login to see this link

I'm using like that...

CyberSEO said
Please refer to the documentation for the correct placeholder names: https://www.cyberseo.net/content-syndicator/#placeholders

%gpt_promt[..]% in your assignment must be fixed as %gpt_prompt[..]%  

%gpt_promt[prompt]% – will be replaced with the result of the prompt processed by the AI model set in Default AI model, allowing this result to be used inside other placeholders with multiple nesting or within shortcodes (without nesting). See the examples below;

fix that line in documentation please. Thank you