Ideogram 4.5 edit with 5 images: only the first is edited, $0.075

On Sume, Ideogram 4.5 edits the first image and treats up to four more as references. Medium costs $0.075 per image; omit aspect_ratio to keep the source shape.

5 min readSume
All posts

On Sume, Ideogram 4.5 (ideogram/ideogram-v4.5) edits the first image you pass and uses up to four more only as references, five images in total. An edit costs the same as a text-to-image call at the same quality: $0.0375 low, $0.075 medium (the default) and $0.275 high.

How the inputs work

The Image API docs say that without input_references Ideogram 4.5 generates from text. With references, it edits the first image and uses up to four more as references. Order therefore matters: the picture you want changed goes first, and style or object references go after it.

An edit that sets no aspect_ratio keeps the shape of the source image, so you do not need to pass the ratio of a photo you are editing. quality takes low, medium or high and defaults to medium; resolution takes 1K or 2K.

Ideogram 4.5 on Sume, as of 2026-10-09
SettingAllowed values or behavior
input_references0 for text-to-image; 1 to 5 for edit (first image edited, up to 4 more as references)
qualitylow, medium (default), high
resolution1K, 2K
aspect_ratioomit on an edit to keep the source shape
Price per image$0.0375 low, $0.075 medium, $0.275 high

Price of an edit session

The fal page for Ideogram V4.5 lists $0.03, $0.06 and $0.22 per image by quality and says size does not change the price (read 2026-10-09). Sume's rows are those numbers times 1.25. The docs list one price per image by quality and no separate line for references, so plan on one image charge per edit.

A typical session is a layout pass at low and a keeper at high: 4 x $0.0375 + $0.275 = $0.425 for four tries and one final. At medium throughout it is 5 x $0.075 = $0.375.

Mistakes to avoid

Do not pass a sixth image; the descriptor allows five, and Sume rejects values outside what a model lists with 400 unsupported_parameter, rather than ignoring them. Do not pass mask_url: the docs document it for ChatGPT Image 2.5 edits, and Ideogram's descriptor does not list it. Read the live descriptor with GET /v1/images/models/ideogram/ideogram-v4.5/endpoints if you are unsure.

A quick test plan

Pick one product photo and three reference images. Run the edit once with the product first and once with a reference first, both at low quality ($0.0375 each, $0.075 for the pair). If the second run edits the wrong picture, you have confirmed the order rule on your own inputs for less than a dime.

Then raise quality only for the order that worked. Keep aspect_ratio unset so the edit keeps the product photo's shape, and set resolution to 1K or 2K for the delivery size you need.

Sources

Related posts

More in Models

All Models posts

Written by Sume