Show HN: Feyn open-sources MultiMatte, a promptable background removal model beating SAM 3 by 34.6% on DIS5K

snyy · hn · 2026-09-10

Feyn released MultiMatte, an open-source background removal model with a key twist: promptability. Unlike typical matting models that keep all foreground, MultiMatte lets you name the object you want to keep in plain language and removes everything else — e.g. keep just the dog in a photo with a dog and a bowl.

Key points:

• Built on Meta's SAM 3 (already concept-promptable); the big change is in masking: alpha mattes (per-pixel opacity) instead of binary masks, better handling hair, fur, and motion blur
• Benchmarks: S-measure jumps from 0.674 to 0.908 on DIS5K (+34.6% relative) and from 0.792 to 0.901 on DUT-OMRON (+13.7%) over SAM 3
• Open source: model on Hugging Face (feyninc/multimatte), usable via the NoBg library (github.com/feyninc/nobg)
• Demo at usefeyn.com/multimatte; training details in the official blog post

Original post →

More from Multimodal

• Adobe Firefly video editor walkthrough: one-sentence Quick Cut turns 32s into 20s — LudovicCreator · 2026-10-26
• Hi3D V1.0 open-sourced with 2M training assets and 90% slicing success rate — kimmonismus · 2026-10-26
• Claude Can Replace Your YouTube Thumbnail Designer in 10 Minutes — thedankoe · 2026-10-26
• Multimodal AI Is Not Useless: A Practitioner Pushes Back on Growing Skepticism — thoma_gu · 2026-10-26
• One prompt, full Blender animation: dev says he never touched Blender once — iamfakhrealam · 2026-10-26
• Dense instance segmentation via Muse Wonder + SAM 3.1 on Meta's Model API — nikhilaravi · 2026-10-26

Multimodal channel →