AI wedding invitation image: quote names and date so they render

To make a wedding invitation image, write every line of text in quotation marks in the prompt, ask for a portrait ratio, and proofread the result for each name, date and place. On Sume, GPT Image 2.5 is the first model to try for lettering, with Ideogram V3 and FLUX.2 flex worth a test (text rendering models). For anything you will print, treat the generated card as artwork and set the final text in a layout tool.

Prompt the card

Name the style, the layout, and each text line with its position. Say that no other text should appear. Use 3:4 or 4:5 for a portrait card; both are listed for the GPT family on Sume (Image API).

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "openai/gpt-image-2.5",
  "prompt": "Wedding invitation card, cream paper, watercolour eucalyptus border. Text, centred: 'Maya and Jonas' large script; 'Saturday 14 June' below; 'Lakeside Pavilion' small capitals. No other text.",
  "aspect_ratio": "3:4",
  "quality": "high"
}''

Proofread every name

Names are the highest-risk text, since a model has no spelling to fall back on. Spell a name out once in the prompt if it is unusual, check the output letter by letter, and have a second person read it. If one line is wrong, fix that region with a masked edit, not a full re-roll (text placement fix).

Print or screen

For a screen invitation, the generated image is enough. For print, the pixel size matters; see the 300 DPI sizes post, and set the text in your layout tool over a text-free artwork version when the type must be perfect (print sizes).