Claude/ChatGPT Prompt to Build a Midjourney LinkedIn Preview Template

What this prompt does

This prompt assigns the AI a senior visual designer role and makes it build a reusable Midjourney prompt for a 1200x627 LinkedIn preview card, with the headline added later in Figma rather than generated. You provide the [article_title], [central_visual], [palette], and [text_margin]. It returns a finished editorial magazine-cover-style prompt, composition rules that keep the named safe margin clear, a typography hint for the Figma headline, explicit palette usage, a negative prompt, and the correct flags — plus a two-line note on where the headline sits.

The structure works because it reserves space for text instead of letting the model render it. Midjourney-rendered text almost always looks broken, so the prompt fixes [text_margin] as a clear zone and sets the bold serif headline in Figma over the top. [central_visual] defines the focal image, [palette] (one dominant colour, one accent) keeps a series looking like one brand, and the flags --ar 1200:627 --style raw --v 6 size it for a LinkedIn card. The negative prompt strips AI-gimmicky shine and embedded gibberish.

When to use it

• You need a LinkedIn preview image that makes people stop scrolling.
• You want a reusable template so a series of posts looks like one brand.
• You need a clear margin reserved for a headline you will add in Figma.
• You learned that model-rendered text looks broken and want to avoid it.
• You want the image sized exactly for a 1200x627 LinkedIn card.
• You need a negative prompt to remove AI-gimmicky shine and gibberish text.

Example output

You get a copy-ready Midjourney prompt in a fenced block — an editorial magazine-cover image at 1200x627 with the safe margin kept clear, a typography hint noting the bold serif headline is added in Figma, one-dominant-one-accent palette use, a negative prompt removing shine and gibberish, and the flags --ar 1200:627 --style raw --v 6 — followed by a two-line note on where the headline sits.

Pro tips

• Set [text_margin] to a real reserved zone, like the left third, so the headline overlay has clean space and the composition respects it.
• Keep [palette] and the margin fixed across a series so your feed reads as one consistent brand.
• Do not let the model render the headline; reserve the margin and set every headline in Figma, since generated text reliably looks broken.
• Use [central_visual] to define a single clear focal image rather than a busy scene that competes with the overlay.
• Lean on the negative prompt to kill AI-gimmicky shine and embedded gibberish, adding terms only if new artefacts show up.
• Trust the --ar 1200:627 flag so the export drops straight into a LinkedIn card without cropping surprises.