Public beta · free during beta

Background removal API for developers.

Drop an image in. Get a transparent PNG out. ~200 ms per call. BiRefNet on Modal GPUs. MIT licensed, self-hostable, 40× cheaper than remove.bg.

bash
curl -X POST "https://useknockout--api.modal.run/remove" \
  -H "Authorization: Bearer $TOKEN" \
  -F "file=@cat.jpg" -o out.png
OriginalOriginal photo
CutoutBackground-removed cutout
200ms
Warm latency, 1024×1024
$0.005
Per image (paid tier)
40×
Cheaper than remove.bg
MIT
Code + model weights

SOTA quality. Real photos.

BiRefNet ranks #1 on DIS5K and HRSOD. We benchmark against remove.bg on hard subjects: hair, fur, glass, group photos.

Hair

Wisps without halos

Closed-form foreground matting cleans soft edges. No fringing, even on stark color contrasts.

Glass

Transparent objects

Cocktail glasses, perfume bottles, water — handled cleanly. The matte preserves real opacity.

Multi-subject

Group photos

Multiple people, overlapping subjects, complex backgrounds. BiRefNet handles them in one pass.

Stickers

Sticker style

iMessage / WhatsApp / Telegram-ready cutouts with a configurable outline width.

Install in your stack.

Official SDKs for Node, React, CLI, and Python. Or hit the REST API directly.

npm i @useknockout/node

import { Knockout } from "@useknockout/node";
const client = new Knockout({ token: process.env.KNOCKOUT_TOKEN });
const png = await client.remove({ file: "./input.jpg" });

Pricing

For reference — the same image on remove.bg is $0.20 at their PAYG rate.

Free

$050 images / month

Best for: Eval, side projects, OSS

  • 50 images / month
  • No card required
  • Public beta token
  • Community support
Start free
Most popular

Pay-as-you-go

$0.005per image

Best for: Side projects, early startups

  • Per-image billing
  • Credits never expire
  • All 20 endpoints
  • Email support
Get token →

Volume

$0.003per image at 100k+/mo

Best for: Production workloads

  • Volume discount
  • 99.9% uptime SLA
  • Slack support
  • Usage dashboard
Talk to us

Enterprise

Customprivate endpoints

Best for: Compliance, BYO-cloud

  • Self-hosted on your VPC
  • Custom SLAs & DPA
  • Dedicated support
  • Private Slack channel
Contact sales

Frequently asked questions

Warm latency is around 200ms for a 1024×1024 image on a Modal L4 GPU. Cold starts add ~3 seconds. Most production calls hit a warm container.