Background removal API and self-hosted open-weight model for image matting
Upload a selfie, get five celebrity matches from 350 public figures
Stage
Not given
Not given
Founded
2022
Not given
Based in
DE
Not given
Pricing model
Freemium
Free
Pricing
Pro Model API: 50 free credits on signup, then credit packs from €10 (100 credits at €0.10 each) to €500 (10,000 credits at €0.05 each), or equivalent monthly subscriptions. Open-weight model: Apache-2.0, free to self-host.
Not given
Key features
Pro Model API with managed REST infrastructure in Frankfurt, Germany
Open-weight Apache-2.0 model for self-hosting on Python, Docker, or macOS
Alpha matte output for precise edge refinement
Refine feature with Keep and Remove brush strokes
Optimized for hair, fur, and complex edges
Native macOS app with batch processing and on-device Apple Silicon inference
GIMP 3.0 plugin for local background removal
Docker support with web UI and headless API
Compares uploaded photo with 350 public figures
Returns five ranked celebrity matches
Face detection and alignment using OpenCV YuNet
Feature vector comparison using cosine similarity
Photos processed in memory, never saved or used for training
Generates shareable downloadable card with match
No account creation required
Supports JPG, PNG, and WebP formats up to 8 MB
Browses 928 licensed reference photos from Wikimedia Commons
Combines multiple reference photos per celebrity into single prototype
What makes it different
Publishes benchmarked results comparing against alternatives including remove.bg with losses included
Same API schema for both managed API and self-hosted open-weight model for easy switching
RAM-only processing with no image storage or training on customer uploads
p95 latency of 1.95 seconds on managed API with EU hosting for GDPR compliance
No subscription lock-in with optional monthly billing or one-time credit packs
No photo storage or training reuse of uploaded images
Transparent methodology with published accuracy limitations and test evidence
Based on real feature comparison, not random results or made-up percentages
Licensed celebrity library with attribution and Wikimedia Commons sources