Ideogram releases BG Remover, a cutout using a generative model rather than segmentation.

Ideogram launches BG Remover, a generative background removal tool that outperforms segmentation models on complex edges for $0.01 per image via its API.

Ideogram completes its graphic ecosystem with BG Remover, an image background removal tool available on all its plans, including the free tier, and via the API. The technical novelty lies less in the function than in the method: whereas almost all market tools (remove.bg, Photoroom, BiRefNet, RemBG, Bria 2.0, Pixelcut) rely on segmentation models that classify each pixel as foreground or background, Ideogram has trained a generative model specific to the task.

The challenge lies at the edges. When a pixel naturally blends foreground and background, segmentation models produce halos, fringing, or color residues in the alpha channel, defects particularly visible on hair, glass, liquids, fine typography, and complex cutouts. The generative approach consists of reconstructing these border pixels as partial transparency associated with the foreground color. The result preserves refractions through glass, fine details, and no longer requires manual retouching in Photoshop.

To explain their advantage, Ideogram's teams highlight their research unit's prior work on reliable text rendering in text-to-image models, which necessitated a granular understanding of counter-forms, fine contours, and antialiasing. This expertise directly transfers to the background removal problem.

The price is set at $0.01 per image, twenty times cheaper than remove.bg at $0.20. The API consists of a POST request to the /v1/remove-background endpoint, which returns a signed URL to the transparent PNG. JPEG, PNG, and WebP are accepted as input. Volume tiers are planned for high-throughput teams, for use cases such as product catalogs, marketplaces, or seasonal refreshes.