Ideogram 4.5

Ideogram 4.5 generates images from text prompts and edits source images using natural language instructions.

You can run it as a text-to-image model by providing only a prompt, or edit up to five source images by passing them in images. Use size="source" to preserve source proportions. If you provide a mask, black pixels are edited and white pixels are preserved.

Billing is per generated image. Source-image edits have separate rates for some quality levels.