AI images for presentations: a cover, illustrations and iconsImage generation with a neural network — the big guide

AI images for presentations: a cover, illustrations and icons in one style

AI writes the text of a deck well. The weakest part of a finished deck is usually its images. Stock photos don’t match each other, the icons come from three different sets, and the cover shows a man in a suit shaking hands with another man in a suit. Generation solves this differently: every image is drawn for a specific slide, and they all come out in one style.

This chapter is only about visuals. If you need the whole deck, from structure to presenting, start with the “AI for presentations” guide.

Which images a deck needs

  • Cover. One strong image on the first slide sets the tone for the whole deck. It can take the full slide, with the title laid on top.
  • Dividers. If the deck has several parts, a simple image in the same style on each section slide tells the audience a new topic has started.
  • Illustrations for key points. Not on every slide — only where the idea is abstract: “growth”, “risk”, “team”. Without an image, that slide is just a list.
  • Icons. Three to five points in a row read better when each has an icon. Generate them in one request, not one by one, so they come out alike.
  • Background. A calm texture or gradient in your brand colours, with text that stays readable on top.

How to keep one style across the deck

The main mistake is describing each image from scratch. The model picks a new style every time, and the slides look like they came from different decks. This order works instead:

  1. Style first, subjects second. Describe the style once: “flat vector illustration, navy and orange, lots of white space, no text”. Repeat that description word for word in every request.
  2. Cover first. Generate it, pick the version that works, and from then on ask for “the same style as the cover”.
  3. Icons as one set. “A set of five icons: growth, team, money, deadline, risk — one style, plain background”.

Try it right here: describe a cover for your topic.

Loading…

Put the text on the slide, not in the image

Image models still garble letters, especially in non-Latin scripts: a title drawn into the image often comes out misspelled, and it can’t be fixed afterwards. It is safer to ask for an image with no text and add the title on top of it, on the slide itself. That way the text stays editable, and you can translate or correct it at the last minute.

The right format for a slide

A standard slide is landscape, 16 : 9 (1280 × 720 px). A full-slide image has to be just as wide, so say so in the prompt: “landscape image, 16 : 9”. A finished image can’t be cropped to another format later, so set the format from the start. For an illustration on half a slide, a square or portrait image works. All sizes in pixels and centimetres are in the “Presentation slide size” chapter.

What’s next

How AI puts text and images together into slides is covered in “A neural network for presentations”. How to choose a layout where images and text don’t compete is in “Slide design and layout with AI”. More on image prompts is in the image generation guide.

Frequently asked questions

Which AI works for presentation images?
Any image generator that understands a style description. The order of work matters more than the model: describe the style once and repeat it in every request. You can generate slide images in Twelver.
How do I keep all the images in one style?
Describe the style in one phrase (technique, 2–3 colours, “no text”) and paste it into every request word for word. Start with the cover, pick the version that works, then ask for “the same style”. Generate icons as one set, not one by one.
Can I ask the AI to write the title right on the image?
You can, but letters — especially in non-Latin scripts — often come out wrong, and they can’t be fixed. It is safer to generate the image with no text and put the title on top of it on the slide.
What size should a slide image be?
For a full-slide image, landscape 16 : 9, like the slide itself (1280 × 720 px). Set the format in the prompt from the start: a finished image can’t be cropped later. For an illustration on half a slide, a square or portrait image works.
Can I make presentation images for free?
You can try without paying: in Twelver, starter tokens arrive on your first visit and signing up adds more. That is enough for a cover and a few illustrations.

In Twelver, slide images are generated in the same conversation where the text is written, and the presentation maker builds the finished deck.

Try it yourself

Everything in this guide runs inside Twelver

One chat for text, images, video, music and voice — no separate services or subscriptions.

Open Twelver chat
Rate your experience