Standard Operating Procedure for generating banner images per blog post using the brand's trained ChatGPT chat. One-part manual process with no companion skill.
Single-part, fully manual process. Output is one finished banner image per blog title, generated in ChatGPT using the brand's established visual style.
This step runs after the blog-post-generator skill has produced the month's blog titles. It covers generating one banner image per blog post (and newsletter, since they share the same banner). It does not cover resizing, logo placement, or client delivery — those are later, separate steps.
The key to keeping this fast is a properly set-up, brand-trained chat in ChatGPT. Once the chat has seen the brand's visual references, individual prompts become a single line.
In the company ChatGPT account, go to the specific brand's Project. Look for a chat named something like "Blog banner image generator" (or similar). Most active brands already have one set up and trained.
If the chat doesn't exist yet for this brand, create it now — it will persist and accumulate context across every future monthly run.
First time only — skip if the chat already exists and has been used before.
Attach visual references before writing any prompt. Use one of the following, in order of preference:
Always select "Create Image" as the action and 16:9 as the image ratio before generating.
For each blog title in the month's content plan, in order:
Repeat for every blog title. The context carries within the same chat — no need to re-reference prompts.
Download each banner immediately after approving it and name it using this convention:
The number corresponds to the blog's order in the content plan for that month. Name at download time — not at the end of the session — so the order is never lost.
blog.N convention has to be applied the moment the file leaves ChatGPT, or order gets lost in a folder of untitled images.