Ideogram: ideogram-4-5/generate

Ideogram 4.5 is an image generation and editing model from Ideogram. It accepts text and image inputs and produces image outputs. Text-to-image generation works at 1K and 2K presets, with native 2K output and supported dimensions reaching roughly 3K. The model supports prompts of up to 10,000 characters and offers four quality modes. Aspect ratios include 1:1, 3:2, 16:9, 9:16, 2:3, 3:4, and 4:5. Editing is a core strength. It edits a source image using up to four additional reference images, optionally with a mask so changes land only where intended—mask-based inpainting preserves white pixels while editing black pixels. The Precise Edit variant applies a natural-language instruction to a source image and restores unchanged pixels from the source, keeping the source dimensions and reducing unwanted visual drift. It handles high-resolution images without downsizing first—demonstrated on a 4,016 by 6,016 source of 24.2 megapixels—and can edit a crop and merge the result back into the original. It also alters stylized lettering in place, translates copy while retaining layout, and supports pose control with a pose skeleton reference and joint angles in the prompt. This makes Ideogram 4.5 suitable for tasks where wording and fidelity matter: in-image copy changes, product campaign variants that preserve a product's appearance across turns, architectural visualization with swapped furnishings and finishes, localization of headlines and labels, and targeted high-resolution edits. It is proprietary, with an open-weight release planned but not yet available.