Bria RMBG 3 — Image Background Removal API
What is Bria RMBG 3?
Bria RMBG 3 is an image background removal model: send one image, get back a transparent PNG — no prompt needed. It is Bria's newest background remover, built for the edges that defeat most tools: flyaway and curly hair, animal fur, sheer fabric, open knits, glass and other transparent objects, and fine detached detail. Rather than a hard binary mask, RMBG 3 produces a non-binary alpha with hundreds of transparency levels, so soft edges and see-through regions blend onto any new background without a cut-out fringe. It is trained entirely on licensed data for safe commercial use.
Key Features
- Transparent PNG with a true alpha channel from one image input.
- Clean cutout edges on hair, fur, glass, sheer fabric and fine detail.
- Non-binary, soft-edged alpha for natural compositing.
preserve_alphakeeps existing transparency in RGBA inputs.- Optional input and output content moderation.
- Licensed-data training for risk-free commercial use.
Best Use Cases
Reach for RMBG 3 when the cutout is the final asset and its edges show: e-commerce product shots with pets, hair or glassware; fashion catalogs with loose hair and sheer fabric; portrait isolation; and ad composites that drop a subject onto a new background. In testing, a long-haired cat shot against a cluttered flower-market stall came back with whiskers, ear tufts and flyaway fur intact and the background fully removed.
Input Tips and Output Quality
RMBG 3 takes no prompt, so quality starts with the input: use a sharp, well-lit, high-resolution image where the subject reads clearly against its background. Keep preserve_alpha on for inputs that already carry transparency. Expect a few seconds per image — RMBG 3 trades RMBG 2.0's sub-second speed for edge quality, so use RMBG 2.0 for interactive previews or high-volume batches.


