The character consistency model works well for single-reference images, but for multi-panel manga workflows I'd love to see a "character library" feature where you can save and recall character appearances across sessions. Also, batch generation for comic pages would be huge for creators producing weekly content.
Ideogram just released Ideogram 4.6, its image generation and editing model for precise, multi-turn edits. You use each result as the source for the next edit, and it generates natively at up to 4K instead of upscaling.
What improved over 4.5:
Exact color: 46% to 83%
Character consistency: 69% to 85%
Product detail: 61% to 86%
Product shots: 66% to 88%
Masked edits: 82% to 90%
Ideogram also says it leaves far less pixel drift outside your edits and holds color more accurately.
What you can do with it:
Multi-turn editing: keep refining one image, step by step
Precise Edit: change just part of an image, with an optional mask
Reference images: use up to five when generating
Four quality modes: from Very low to High, with native 4K output on higher modes
API: a generate endpoint and a Precise Edit endpoint
On Rapidata’s Multi-step Image Edit leaderboard (November 17, 2026 snapshot), 4.6 ranks first at 1,892 ±32, ahead of 4.5 (1,866) and Seedream 5.5 (1,805). Ideogram notes that this comes from one configuration and may not apply to every workflow.
API pricing per image runs from $0.008 (Very low, 2K) up to $0.20 for High 4K generation, and High 4K edits cost $0.60. It’s available in the Image app’s Model picker when it has reached your account.
Good fit if you iterate on brand visuals, product shots or characters and need edits that keep colors and details intact. Ideogram suggests checking spelling and placement after editing text in an image, so check that too.
Learn more: Ideogram 4.6