Ideogram released a new version of its image editing tool called Ideogram 4.5, and the pitch is simple: change one part of a photo and leave everything else exactly as it was. Swap a shirt color, move a chair, fix a torn corner of an old photo, and the rest of the image should not shift at all.
That sounds like a small thing, but it has been the hardest problem in AI photo editing. Ask a model to make one change, then another, then another, and small errors pile up. Faces drift, colors shift, backgrounds blur. Ideogram says 4.5 holds steady even after repeated edits to the same image, which matters most for anyone doing a long string of small tweaks rather than one-off edits.
This is not a problem Ideogram is solving alone. OpenAI's image tool and Google's Gemini based image model, nicknamed Nano Banana, have both been chasing the same fix. Google's latest version specifically markets its ability to keep faces and objects steady across edits. The fact that three separate companies are racing toward the exact same fix tells you this was the real bottleneck holding back business use of AI image editing, not image quality itself.
Ideogram is a small player in this fight. It was started in Toronto in 2022 by a group of researchers who had previously worked on Google's own image generation research. The company has raised under 100 million dollars total, a rounding error compared to what OpenAI or Google spend in a year. Yet it keeps releasing models that hold their own against those two giants, which says more about how specialized a problem image editing has become than it does about company size.
The pricing is built for volume use. Four quality tiers run from under a penny to about 22 cents per image, all producing sharp, high resolution output. That price range matters for any business that edits large batches of photos regularly, such as an online retailer updating thousands of product photos for a new season, or a real estate company generating interior staging images. At those prices, running thousands of edits a month costs less than a single afternoon of a photo retoucher's time.
Ideogram also plans to release an open version of the model that businesses can run on their own servers instead of paying per image. For a retailer or marketing team editing tens of thousands of photos a month, that changes the cost math substantially, since the per-image fee disappears once you own the infrastructure.
There is a flip side worth sitting with. As editing gets cleaner and harder to spot, verifying whether a photo has been altered gets harder too. That matters for anything where a photo is supposed to prove something happened: insurance damage claims, real estate listings, product condition reports. The same precision that makes this tool useful for legitimate editing also makes convincing fakes easier to produce.
The near term effect for most businesses is straightforward. Routine photo editing work, product photo touch ups, catalog consistency, simple restorations, is getting cheap and fast enough that doing it in house with one of these tools becomes the default, not the exception.