What is a background separation API?

A background separation API is a web service that separates the subject of an image from its background and returns both as independent layers. You send a source image, and the API delivers two outputs: the subject with an alpha channel (transparent PNG cutout) and a fully inpainted background with the subject erased. Used by ad creative teams, photo editors, AR/virtual try-on products, and video pipelines to enable layer-aware editing workflows.

How is background separation different from background removal?

Background Removal returns one output — the subject with a transparent background, discarding the background content entirely. Background Separation returns two outputs — the subject (same as removal) plus a fully reconstructed background with the subject erased via context-aware inpainting. Separation is for workflows where you want to keep or edit the background too; Removal is for workflows where you just want to discard it.

When should I use Background Separation vs Background Removal?

Use Background Removal when you only need the subject — you'll discard the background (e.g., transparent PNGs for marketplace listings, catalog cutouts). Use Background Separation when you want to edit, composite, or replace the background without losing its content (e.g., AR try-on, ad variations across backgrounds, virtual staging). Separation returns two usable layers per call; Removal returns one. If your workflow ends with a transparent PNG, use Removal. If your workflow includes independent foreground/background editing, use Separation.

What outputs do I receive?

Two images per call: (1) the foreground — subject isolated with alpha channel, transparent PNG format; (2) the background — subject silhouette filled with context-aware inpainting matching surrounding scene content. Both are production-ready for compositing, editing, or replacement workflows.

How does AI background separation work?

The API runs subject extraction (AI segmentation with edge refinement) in parallel with context-aware inpainting (filling the subject's silhouette in the background with plausible scene content). The two outputs are generated together so edge alignment is exact — the background's filled region matches the subject's silhouette precisely.

Does it handle hair, fur, and complex edges?

Yes. Fine details like hair, fur, fabric, transparent objects, and complex outlines are handled with edge refinement. The subject's alpha channel preserves these details, and the background inpainting fills around them naturally — no halos, no color fringing, no obvious artifacts.

What happens with multiple subjects in the image?

Images with multiple subjects are handled by extracting all detected subjects as a single foreground layer, with the background inpainted around them together. For example, a group photo returns one alpha cutout containing all people, and a background with everyone's silhouette inpainted. If you need per-subject separation (each subject as its own independent layer), you'll need multiple API calls with different inputs or downstream processing — contact us for advanced pipelines.

Does the subject stay pixel-identical in the output?

Yes. The foreground output contains the subject's original pixels unchanged — only the alpha channel is added to isolate it from the background. No retouching, no re-rendering, no modification to the subject's content. Critical for product photography and editorial work where subject fidelity matters. The background layer is generated content (inpainted to fill the silhouette), but the foreground stays true to the source.

Do I need rights to the image I'm separating?

Yes. You need rights to the source image — your own photography, licensed stock, or images you have permission to modify. The foreground output is a direct extraction from your source, so its rights follow the source. The inpainted background includes AI-generated content (filling the subject's silhouette), but that generated content is derivative of your source image and inherits the same rights. For commercial use, prefer your own photography or properly licensed stock.

How much does the background separation API cost?

$0.20 per image with simple fixed pricing. Includes both foreground and background outputs per call. Sign up and get $10 in free GPU credits to try it out. Contact us for volume pricing on bulk workloads.

What happens if the API can't detect a clear subject?

The API returns a structured response for every call. If no clear subject is detected (e.g., abstract imagery, landscape-only shots, heavily cluttered scenes with no focal element), you'll receive an error code with a descriptive message so your integration can handle the case — surface to the user, fall back, or retry with a different crop. For transient infrastructure issues, internal retries are included in the price.

Can I process images in bulk?

Yes. Batch endpoints with async webhook delivery. Process thousands of images per job — ideal for campaign rollouts, AR library preparation, video pipeline feeds, and catalog variation generation.

How fast is the background separation API?

Typical response times are under 5 seconds for standard resolution images. Built for production workloads with consistent latency at scale.

What image formats are supported?

JPEG, PNG, and WebP. Output format depends on the solution.

Typical response times are under 5 seconds for standard resolution images.

How does pricing work?

Simple fixed per-call pricing. No GPU costs, no infrastructure to maintain.

Live

People ImageryOn-Model / FashionProduct Imagery

Edit foreground and background independently

Background Separation API

Separate subject from background and get both layers back as independent outputs. Edit, composite, or replace foreground and background separately with one API call.

Get API access

Start with $10 in free GPU credits on signup — no commitment

Before

After

Rated 5/5 by early access developers

Trusted by teams shipping AI-powered products

BetterPic Case Study

ImagehostVisualIDMarkedBetterStudio

<5s

Average processing time

$0.20

Per image, fixed pricing

4096px

Max resolution supported

2 layers

Foreground + background, both returned

How it works

No pipeline to build, no models to configure.

Send your image

Pull a source image with a clear subject and background. JPEG, PNG, or WebP up to 4096x4096 pixels.

AI separates the layers

Runflow's pipeline extracts the subject with an alpha channel and generates a fully inpainted background in parallel — both layers produced in one pass.

Get both outputs

Receive the foreground (transparent PNG with alpha) and the background (inpainted to fill the subject's silhouette). Each call is $0.20/image and delivers both.

Subject and background, both returned

One API call returns two production-ready outputs — a clean subject with alpha channel, and a fully inpainted background with the subject erased. No more running two APIs or stitching outputs together.

Foreground layer — subject isolated with alpha channel

Background layer — subject silhouette filled via context-aware inpainting

Both layers delivered in a single API call

Edit foreground and background independently

Once separated, each layer is its own asset. Change the background without touching the subject. Retouch the subject without re-rendering the scene. Composite both into new compositions with precise, layer-level control.

Retouch subject without affecting background

Swap or edit background without losing foreground detail

Clean seams — no halos, no edge artifacts

Production-grade edge handling

Handles the hard edges automatically — hair, fur, fabric, transparent objects, complex outlines. The subject's alpha channel preserves fine details, and the background inpainting fills around them with scene-matched content.

Hair, fur, and fine-detail edges handled cleanly

Transparent and semi-transparent objects supported

Background inpainting matches scene lighting and texture

Built for compositing workflows at scale

Designed for teams building layer-aware pipelines — ad variation generators, virtual try-on, AR products, video compositing. Integrate as a single step; downstream systems work with both layers independently.

Two usable assets per call — downstream pipelines don't need to do more

Batch processing with async webhook delivery

JPEG, PNG, and WebP input/output up to 4096x4096px

Background separation for every compositing workflow

From ad campaign variations to AR try-on. Two usable layers per call, ready for independent editing.

Ad Creative & Marketing Teams

Generate campaign variations by swapping backgrounds independently of the subject. One shoot, unlimited creative across markets, placements, and seasons. Think: Meta Ads, Google Shopping, TikTok Ads.

Ecommerce & Virtual Try-On

Power virtual try-on and AR products. Separated layers let you place users into any scene, or scene elements into any user photo. Think: Amazon Try, IKEA Place, Warby Parker try-on.

Photo Editing Apps

Add advanced compositing features to your photo editor. Let users edit subject and background separately with one API call — no manual masking required. Think: Photoroom, Canva, Facetune.

Video & Motion Graphics Pipelines

Per-frame subject and background separation for compositing, VFX, and motion design. Batch-process entire sequences with consistent separation across frames.

AI Image Platforms

Layer-aware post-processing before delivery. Subject and background can be polished independently — different models for each — giving more control over final output quality.

Virtual Staging & Real Estate

Separate occupants or staging props from room backgrounds for independent editing. Stage empty rooms or de-stage furnished ones without losing either layer. Think: Zillow, Airbnb, Compass.

“

Working with Runflow changed how we build. Our developers can integrate a new AI feature by calling an endpoint — they don’t need machine learning expertise, infrastructure knowledge, or DevOps skills. We just focus on making the best product, and Runflow handles everything underneath.

Thibaut Hennau

CEO, BetterPic

Simple, per-image pricing

One price per separation. Both layers included. No GPU fees, no infrastructure to manage.

$0.20/image

Fixed pricing for background separation.

One API call returns two production-ready layers. No setup fees, no surcharges, no per-layer add-ons.

$10 in free GPU credits on signup

Two outputs per call — foreground + background

Subject isolated with alpha channel, background inpainted

Handles hair, fur, and complex edges

Up to 4096x4096px resolution

Batch processing with async webhooks

Under 5 seconds processing time

Get API access

Build vs. buy

Build your own

Weeks to ship

Subject segmentation + edge refinement + inpainting model + layer compositing + batching

Runflow API

$0.20/image

One API call, two layer outputs, under 5 seconds

Processing large volumes?Contact for volume pricing

Your data stays yours

Images are processed in isolated environments with automatic cleanup. Zero data retention by default.

SOC 2 Type II

ISO 27001· In progress (Q3 2026)

GDPR

SOC 2 Type II compliant.

AES-256 encryptionIsolated processingEU data residency available

Visit Trust Center

Frequently asked questions

Edit subject and background. Independently. At scale.

One API call. $0.20/image. Two outputs per call — foreground with alpha, background inpainted. Start with $10 in free GPU credits on signup.

Get API access Browse all solutions

First-in-line access: onboarding slots limited this month

Background Separation API

How it works

Send your image

AI separates the layers

Get both outputs

Subject and background, both returned

Edit foreground and background independently

Production-grade edge handling

Built for compositing workflows at scale

Background separation for every compositing workflow

Ad Creative & Marketing Teams

Ecommerce & Virtual Try-On

Photo Editing Apps

Video & Motion Graphics Pipelines

AI Image Platforms

Virtual Staging & Real Estate

Simple, per-image pricing

Build vs. buy

Your data stays yours

Frequently asked questions

What is a background separation API?

How is background separation different from background removal?

When should I use Background Separation vs Background Removal?

What outputs do I receive?

How does AI background separation work?

Does it handle hair, fur, and complex edges?

What happens with multiple subjects in the image?

Does the subject stay pixel-identical in the output?

Do I need rights to the image I'm separating?

How much does the background separation API cost?

What happens if the API can't detect a clear subject?

Can I process images in bulk?

How fast is the API?

Edit subject and background. Independently. At scale.