Skip to content

AI posters, banners and covers with text

To make an ad poster, banner or cover with text in ClickBait, open the Photo section, describe the image and put the headline in quotes. If the ad is built around a product, attach its photo as a reference. GPT Image and Seedream handle lettering well, and Nano Banana is handy for editing a finished version from a photo. Each run gives up to four variants, and the frame can be square, vertical for stories or a horizontal banner.

Updated:

Models for posters and banners

Ad posters and banners: models in ClickBait
VersionDeveloperModesResolutionAspect ratiosReferencesPromptFeaturesPrice from
Kling O3 ImageKlingtext to image, edit from photos1K, 2K, 4K1:1, 9:16, 16:9, 21:9, 2:3, 3:2, 3:4, 4:3up to 10 imagesup to 4,000 charactersfrom 2 tokens
Nano Banana 2Googletext to image, edit from photos1K, 2K, 4K1:1, 9:16, 16:9, 21:9, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4up to 14 imagesup to 4,000 charactersfrom 6 tokens
Nano Banana ProGoogletext to image, edit from photos1K, 2K, 4K1:1, 9:16, 16:9, 21:9, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4up to 4 imagesup to 4,000 charactersfrom 12 tokens
Nano BananaGoogletext to image, edit from photos1:1, 9:16, 16:9, 21:9, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4up to 4 imagesup to 4,000 charactersfrom 4 tokens
GPT Image 2.5 FlareOpenAItext to image, edit from photos1K, 2K, 4K, custom size1:1, 9:16, 16:9, 3:4, 4:3, 2:3, 3:2, 4:5, 5:4, 21:9up to 16 imagesup to 32,000 characters5 quality levelsfrom 2 tokens
GPT Image 2.5 SunburstOpenAItext to image, edit from photos1K, 2K, 4K, custom size1:1, 9:16, 16:9, 3:4, 4:3, 2:3, 3:2, 4:5, 5:4, 21:9up to 16 imagesup to 32,000 characters5 quality levelsfrom 2 tokens
GPT Image 2OpenAItext to image, edit from photos1K, 2K, 4K, custom size1:1, 9:16, 16:9, 3:4, 4:3, 2:3, 3:2, 4:5, 5:4, 21:9up to 16 imagesup to 4,000 characters3 quality levelsfrom 2 tokens
GPT Image 1.5OpenAItext to image, edit from photos1:1, 2:3, 3:2up to 1 imagesup to 4,000 charactersgood at rendering text in imagesfrom 2 tokens
GPT-4o ImageOpenAItext to image, edit from photos1:1, 2:3, 3:2up to 2 imagesup to 4,000 charactersfrom 2 tokens
Seedream 5.0 ProByteDancetext to image, edit from photos2K1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 21:9up to 10 imagesup to 4,000 charactersfrom 8 tokens
Seedream 5.0 LiteByteDancetext to image, edit from photos2K, custom size1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 21:9up to 4 imagesup to 4,000 charactersfrom 3 tokens
Seedream 4.5ByteDancetext to image, edit from photos2K, 4K, custom size1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 21:9up to 8 imagesup to 4,000 charactersfrom 4 tokens
Seedream 4ByteDancetext to image, edit from photos2K, 4K1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 21:9up to 8 imagesup to 4,000 charactersfrom 4 tokens
FLUX.2Black Forest Labstext to image, edit from photos1K, 2K1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3up to 8 imagesup to 4,000 charactersfrom 3 tokens
Flux KontextBlack Forest Labstext to image, edit from photos1:1, 9:16, 16:9, 3:4, 4:3, 21:9, 9:21up to 1 imagesup to 4,000 charactersfrom 4 tokens
Wan 2.7 ImageAlibabatext to image, edit from photos1:1, 9:16, 16:9, 3:4, 4:3up to 4 imagesup to 4,000 charactersfrom 2 tokens
Z-ImageAlibabatext to image1:1, 9:16, 16:9, 3:4, 4:3no referencesup to 4,000 charactersfrom 2 tokens

How to make an ad poster with AI

  1. Describe the poster

    Say what is in the frame, where the product sits and what the text says. Type the headline in quotes, word for word, and attach the product photo as a reference.

  2. Pick a model and a frame

    Seedream 4.5 is selected by default, and for tricky lettering GPT Image is worth a try too. Choose the aspect ratio for the platform where the ad will run.

  3. Run a series

    Ask for one to four variants at a time. The run button shows the token total before you start, and the best variant can be refined later.

Which models handle text in images

For lettering, go with GPT Image or Seedream: both render headlines, slogans and package labels neatly. GPT Image 2 and 2.5 take up to 16 references, so a single poster can combine the product, your logo and a style sample.

Reach for Nano Banana when you already have a base: upload a photo or an earlier version of the poster and ask it to change the background, the color of a badge or the words in the headline. FLUX.2 suits photorealistic scenes built from several references, while Z-Image works from text only and ignores uploaded photos.

Models render a short headline more reliably than a paragraph of small print. Long promo terms or a footnote are easier to add as a separate layer in a graphics editor.

What to put in a poster prompt

  • The main subject and where it sits, for example "a lemonade can on the right, on a wet stone".
  • The headline in quotes, exactly as it should appear in the image. The shorter the phrase, the cleaner the letters.
  • The text hierarchy: what is large, what is smaller and where it goes. For instance, the headline at the top, a subheading below it, a call-to-action button at the bottom.
  • A palette of two or three colors: "dark blue background, yellow letters, white badge".
  • The style: photorealism, flat illustration, 3D, retro poster or minimalism.
  • The format and room for text: "vertical poster, clean background at the top for the headline". Set the same format in the settings too, since they control the image size.

Example prompt for a banner

Ad banner for a coffee shop. On the right, a glass of iced coffee covered in condensation, soft studio light. On the left, a large headline "ICED COFFEE", below it in small type "new summer menu", and at the bottom a yellow badge that says "Try it today". Palette: dark brown and cream, minimalist style, no extra objects.

This prompt gives the model everything it needs: the subject and its place, three levels of text from top to bottom, the colors and the style. The model tries to copy quoted words letter for letter, so catch typos while they are still in the prompt. For a banner like this, pick 16:9 in the settings.

Aspect ratios for ad platforms

Social feeds take square 1:1 and vertical 4:5, stories and other vertical screens take 9:16, and a website banner or a video thumbnail takes horizontal 16:9.

Each model has its own set of ratios. Square is available everywhere, and 9:16 and 16:9 almost everywhere; the exceptions are GPT Image 1.5 and GPT-4o Image, which offer only 1:1, 2:3 and 3:2. Vertical 4:5 comes from Nano Banana and GPT Image 2 and 2.5. An exact size in pixels can be entered on GPT Image 2 and 2.5, Seedream 4.5 and Seedream 5.0 Lite.

If a platform caps the file size, a finished image has a "Compress" button that makes the file lighter without a new generation.

A series of variants and refining the best one

One run gives up to four variants of the same prompt. When you test creatives, change one thing between runs: the headline, the background color or the product angle. That way you can see which change worked.

Open the variant you like and click "Use as a reference for photo": the image goes into the composer references. Then describe the edit, such as replacing the headline with "SALE ALL WEEK" or making the background lighter. A model that edits from photos fits this job, for example Nano Banana or GPT Image.

A finished poster can also be animated: the "To video" button puts the image into the Video section as the first frame.

Logos and brands that are not yours

Do not use other companies' logos, names or brand style in ads without the rights holder's permission. A model can draw a recognizable mark, but that does not give you the right to use it, and the ad platform may reject the creative.

Add your own logo, product and packaging as references: the model then sees what they actually look like and has less to guess.

Where to find it in ClickBait

Ad posters and banners lives in ClickBait under the Photo section. Sign up, pick a version from the model list and describe what you want to get.

The token price is shown on the run button before you start. If a generation fails, the tokens go back to your balance automatically.

New accounts created with Google or email get a welcome bonus of 10 tokens. These are promo tokens and expire after 14 days.

Models for this task

  • Kling

    video 3-15 s; matches the source video, up to 4K

    images up to 4K

    up to 10 references

    from 2 tokens

  • Nano Banana

    images up to 4K

    up to 14 references

    from 4 tokens

  • GPT Image

    images up to 4K

    up to 16 references

    from 2 tokens

  • Seedream

    images up to 4K

    up to 10 references

    from 3 tokens

  • FLUX

    images up to 2K

    up to 8 references

    from 3 tokens

  • WAN

    video 2-30 s, up to 1080p

    up to 4 references

    from 2 tokens

Other tools

Frequently asked questions

Which AI is best at text in images?

For a poster with lettering, start with GPT Image or Seedream: both write headlines and short slogans well. If you need to fix a finished image or build a poster around a product photo, Nano Banana is the easier choice.

How do I make a banner with my product photo?

Upload the product photo as a reference, describe the scene and put the headline in quotes. Every model in the table accepts references except Z-Image, which draws from text only.

Can the text be in a language other than English?

Yes, type the quoted text in the language you need. Short words in a large font come out cleaner than a long line of small type. Check the letters on the result, and if one is off, run it again or fix the text from a reference.

Which aspect ratio fits a banner or a cover?

1:1 or 4:5 for feeds, 9:16 for stories, 16:9 for a website banner or a video thumbnail. The table above shows which ratios each model supports.

How do I fix the text on a finished poster?

Open the image, click "Use as a reference for photo" and write what to change: the words, a color or the background. Nano Banana and GPT Image work well for this kind of edit.

Can I put another company's logo on a banner?

Not without the rights holder's permission: someone else's trademark in an ad can infringe the owner's rights, and the platform may reject the creative. Add your own logo and product as references.