Generate a new image from text, or edit an uploaded image with precise instructions.
Describe the image or the exact edit. Put words you want rendered in quotes. Maximum 5000 characters.
Low is fastest and cheapest, Medium is the default, High is best for final typography and fine detail.
Choose 1–8 variants. Each generated image is billed separately.
Sign in to spend tokens
Prompt is required
Your generated images will appear here
Ideogram 4.5 generates design-ready images with crisp typography and precise layouts, and edits existing visuals with a prompt while keeping the rest of the frame intact. On Prizmad it runs on your normal token balance — write a prompt, pick a quality tier, generate.
Ideogram 4.5 is the next generation of Ideogram's design-focused image model. It renders posters, ad creatives, promo graphics, and product visuals where typography and layout matter as much as the imagery — prompt the exact words, composition, and type style you want and it draws them with design-grade precision.
It works in two modes. Text to Image creates a new visual from a prompt, with three quality tiers (Low for fast drafts, Medium as the default, High for final typography and fine detail), standard size presets, optional prompt expansion, up to eight images per run, and a seed for repeatable results. Edit Image takes an uploaded image plus a prompt and changes only what you ask, preserving the rest of the frame — with a High precision option that restores unchanged pixels for product photos and layouts.
On Prizmad, Ideogram 4.5 uses the same token balance as AI avatars, video, voiceover, captions, and music. No separate account, no API key — every image is yours to use commercially.
Social ads and posters where the hook — "7 DAY GLOW", "FRESH START" — is rendered directly in the image with clean, readable type.
Polished launch images and campaign posters with the product centered, brand text placed precisely, and photoreal lighting.
Swap a background, add a reflection, or restyle a shot while keeping the product, label, and typography exactly as they were.
App store promos, sale banners, and social graphics with sharp UI-inspired layouts and balanced negative space.
Regenerate one concept across square, 9:16, and 16:9 presets so a single idea covers feed, Stories, and display placements.
Renders headlines, subtitles, and layout text sharply inside the image — quote the exact wording in your prompt and describe the type style you want.
Follows composition instructions — centered product, balanced negative space, editorial layouts — for design-ready posters and promo graphics.
Upload an image and describe the change. Only what you ask for changes; the rest of the frame, including labels and typography, is preserved.
Switch on High precision to restore unchanged pixels, keeping product packaging, logos, and fine detail intact through the edit.
Low for quick drafts, Medium for everyday work, High for final assets where typography and fine detail need maximum polish.
Generate up to eight images per run — each billed separately — and set a seed to reproduce a composition you have already approved.
Choose a mode: Text to Image to create a new visual, or Edit Image to change an uploaded one.
Write a prompt (up to 5,000 characters) describing the image or the exact edit. Quote the words you want rendered — for example: a poster with the headline "MORNING RITUAL" in bold serif type.
For Edit Image, upload the image to edit first, then up to four reference images (JPG, PNG, or WebP). The first image is the source; the rest guide the result.
Pick a quality tier: Low for drafts, Medium (the default) for most work, High when the final typography and detail matter most.
In Text to Image, choose a size preset — Square HD, Portrait 4:3, Landscape 4:3, Portrait 9:16, or Landscape 16:9 — and set the number of images (1–8; each is billed separately).
Optional: in Edit Image, turn on High precision to restore unchanged pixels; in Text to Image, toggle prompt expansion for richer creative direction and set a seed for repeatable results.
Click Generate. Results typically arrive in about 1–3 minutes and are saved to your asset library.
Ideogram 4.5 is listed from 6 tokens per image in Low quality, 12 in Medium, and 44 in High. Each image in a multi-image run is billed separately. Tokens come from your token balance.
Edit Image runs are billed per generated image by quality tier, so a quick cleanup costs less than a full high-quality rework. Size does not change the price.
Ideogram 4.5 is unlocked on the same token balance you use for AI avatars, video, voiceover, and music — no separate account, no API key setup, no extra billing.
No — each generation consumes tokens from your Prizmad balance, from 6 tokens per image in Low quality, 12 in Medium, and 44 in High. Each image in a multi-image run is billed separately. There is no separate account or API key needed.
Yes. Images you generate with Ideogram 4.5 on Prizmad are yours to use commercially — in paid ads, on landing pages, in listings, and in printed materials.
Text to Image creates a new visual from your prompt. Edit Image takes an uploaded image plus a prompt and changes only what you ask, preserving the rest of the frame. Edit Image keeps the source image's dimensions.
Turn on High precision in Edit Image. It restores unchanged pixels after the edit, so packaging, logos, and fine typography stay exactly as they were. Then review the result before publishing.
Put the exact wording in quotes inside your prompt and describe the typography and placement — for example: 'headline "GLOW DAILY" in crisp bold type, centered above the bottle'. For final assets, choose the High quality tier.
Typically about 1–3 minutes. Low quality is quickest for drafts; High takes longer but delivers the best typography and detail. Generations run asynchronously and land in your asset library when done.
Yes — set a seed in the advanced settings. Using the same seed with the same prompt and settings produces repeatable results, useful for iterating on a composition you have already approved.
Published 2026-10-02 · Last updated 2026-10-02