Ideogram 4.5 aims to keep the image intact between edits.

Ideogram 4.5 targets the drift that builds up across successive edits to the same image. The model is designed to preserve composition, colors, and details between changes, with workflows spanning product photography, typography, architecture, and photo restoration.

Editing an image without rebuilding it every time

Change a product's color, adjust its lighting, modify some text, then work on another detail. After several passes, image editing models tend to gradually alter areas that were never part of the instructions.

Ideogram 4.5 specifically targets that accumulation. The model is designed to reduce unintended shifts in color, texture, and detail between operations so the image remains stable throughout an editing session.

That makes multi-turn workflows more practical: instead of requesting several transformations in a single prompt, each change can be isolated and added to the previous one. A comparison across 14 successive edits

Ideogram compares its model with GPT Image 2.5 Sunburst, Nano Banana Pro, and Nano Banana 2 across a sequence of up to 14 edits.

The company says competing models accumulate more artifacts and unintended changes as the sequence progresses, while Ideogram 4.5 retains the original image more closely. The comparison was produced by Ideogram and is not an independent benchmark.

The demonstrations cover several situations where that stability matters: changing a product's color without altering its shape, relighting a scene while preserving its composition, or progressively restoring an old photograph by separating damage removal, colorization, and enhancement into individual steps. Text can change without rebuilding the design

Typography remains another focus. Ideogram 4.5 can modify stylized text directly inside a composition or translate it into another language while attempting to preserve the existing design.

For product photography, the same precision can be used to create multiple colorways or environments while keeping the product itself consistent across images.

The approach also extends to interiors and architecture. Furniture, materials, and palettes can be replaced progressively from an existing photograph while the geometry of the space remains unchanged. Working directly on a high-resolution region

Editing does not necessarily require downsizing the entire image. Ideogram demonstrates the feature on a 4,016 × 6,016 image, or 24.2 megapixels, while modifying only a region around a handbag.

The model preserves the boundaries of the edited area so it can be stitched back into the original file. The workflow is intended for localized changes on images used for close-ups or large-format printing.

The Ideogram API documentation also supports prompt-based image editing, with up to ten images supplied to a single request. Four quality modes, with open weights coming later

Ideogram 4.5 operates at native 2K resolution and comes in four quality modes. Pricing ranges from 0.8 cents to 22 cents per image depending on the selected mode.

Two API endpoints accompany the model: Precise Edit, which keeps the dimensions of the input image, and Generate + Edit, which combines generation and editing.

The model is available through Ideogram, its API, and launch partners including Krea, Leonardo AI, Runway, Pika, fal, Picsart, and LMArena. Ideogram also says open weights are coming soon, without providing a release date.