Ideogram 4.5 Edit

Ideogram 4.5 Edit: Precise AI Image Editing (Image-to-Image)

What is Ideogram 4.5 Edit?

Ideogram 4.5 Edit is a precise image-to-image editing model from Ideogram, the lab known for the most accurate in-image text rendering. You upload a photo and describe the change in plain language; the model applies only that change and copies every pixel you did not ask to touch straight from the source, returning the result at the image's exact original width and height. An optional black-and-white mask confines the edit to a region, and up to four reference images (three alongside a mask) let you bring in a product, logo, pattern, or style. It is built for surgical, multi-turn editing where preserving the rest of the picture matters.

Key Features

• •Prompt-driven precise edits that keep everything you did not ask to change identical
• •Optional B/W mask (black edits, white keeps) at the image's own size
• •Up to 4 reference images (3 with a mask) for product, logo, pattern, or style transfer
• •Four quality and price tiers: very_low, low, medium, high
• •Output always matches the input's exact dimensions, so edited crops stitch back seamlessly
• •Reproducible seeds and optional post-generation copyright detection

Best Use Cases

Add or change legible text on packaging, signage, and products — Ideogram's signature strength — while the rest of the frame stays pixel-identical. Recolor an object, swap a material, or restore a region with a mask. Drop a logo or pattern onto a product with reference images, or restyle a whole image while keeping its lettering sharp. In testing, a plain maple skateboard received a chrome "RAMPWORKS" wordmark with the trucks, wheels, and ground untouched; a masked pass recolored only the deck to candy-apple-red; and a quality=high pass turned the frame into a halftone screen-print poster with the text still crisp.

Prompt Tips and Output Quality

Put the exact words you want in quotes and name the lettering style. Describe the change itself, not the mask, and always state what to keep. Use medium for everyday edits and high for print-grade detail or heavy stylizations. Supply a same-size mask to confine a change, and reference images to bring in an outside design.

How to Use Ideogram 4.5 Edit

Ideogram 4.5 Edit changes exactly what you describe and leaves the rest of your image untouched. Start with a clear source image, then write an instruction that names the change and states what to preserve. Because the output returns at the source's exact size, it suits multi-turn refinement and print-ready work.

How to prompt

• •Describe the edit itself ("recolour the deck to candy-apple-red"), not the mask.
• •For text, put the exact words in quotes and name the style ("bold chrome speed-script").
• •Always list what to keep ("keep the trucks, wheels and background unchanged").
• •Add a same-size black-and-white mask (black = region to edit) to confine a change.
• •Add up to four reference images (three with a mask) to bring in a product, logo, pattern, or style.

Recommended Settings

• •quality: medium for everyday edits (balanced, ~30s, lowest sensible cost). high for print-grade detail and heavy restyles where text must stay legible. very_low / low for quick drafts.
• •mask: optional; must match the image size and contain both black and white areas.
• •reference_images: up to 4 (3 when a mask is also sent).
• •seed: set a value to reproduce a result; omit for variety.
• •enable_copyright_detection: off unless you need flagged outputs withheld.

Example edits to try

Each example below ran end to end at the settings shown; the deck, trucks, wheels, and ground carried over from one source image.

• •Precise text edit (default). Prompt: "Paint the bold wordmark 'RAMPWORKS' in a 1970s chrome speed-script down the centre of the skateboard deck, with a smaller 'EST. 84' beneath it; keep the grip tape, trucks, wheels and concrete ground unchanged". The chrome lettering rendered cleanly on the wood with everything else identical. Tip: put the exact words in quotes and name the lettering style.
• •Reference-guided design. Prompt: "Apply the flame pattern from the reference image as the full graphic across the skateboard deck, following the deck shape; keep the trucks, wheels and concrete unchanged". The reference pattern wrapped the deck while the hardware stayed put. Tip: upload the design as a reference image and say where to apply it.
• •Mask-limited recolour. Prompt: "Recolour the skateboard deck to a deep candy-apple-red gloss finish; keep the grip tape, trucks, wheels and concrete ground exactly the same". Only the masked deck changed colour. Tip: black marks the region to edit on a same-size mask, and still list what to keep.
• •High-quality restyle. Prompt: "Restyle the whole image as a bold two-colour screen-printed poster in black and safety-orange with halftone shading; render the wordmark 'RAMPWORKS' large across the deck, kept sharp and legible". At quality=high the frame became a halftone poster with the text still crisp. Tip: use high for heavy stylizations where lettering must survive.

Reference-guided edits

The model takes up to four reference images (three alongside a mask). Address each one by what it is — "the flame pattern from the reference image", "the logo in the second reference".

Try also: "add a wax-stamped monogram from the reference to the tail", "swap the wheels to translucent amber, keep the deck", "restore the faded grip tape and sharpen the wood grain".

Does it keep the rest of the image unchanged?

Yes — unmodified pixels are copied from the source, and the output comes back at the original width and height.

Do I need a mask?

No, it is optional. Add a same-size black-and-white mask (black marks the region to edit) to confine the change.

How many reference images can I use?

Up to four, or three when you also send a mask.

What output size do I get?

The input's exact dimensions; very large images are scaled proportionally to fit and returned as rendered.

Which quality tier should I choose?

medium by default, high for print-grade detail, and very_low or low for quick drafts.

Can it render text inside images?

Yes — Ideogram leads on legible, correctly spelled in-image text. Put the words in quotes and describe the typographic style.