What is a reference-based inpainting API?

A reference-based inpainting API is a web service that fills a masked region of one image using a second image as visual guide. You provide a target photo, a black-and-white mask marking the region to fill, and a reference image whose content drives the fill. Used by fashion brands, product photographers, ad creative teams, and AI image platforms to swap fabric patterns, fix product defects, and generate creative variations with predictable, style-matched results.

How is reference-based inpainting different from prompt-based inpainting?

Prompt-based inpainting describes the desired fill in words. Reference-based inpainting shows the AI by example. The result is more predictable and visually consistent with your intent, there's no ambiguity from interpreting language, just direct visual guidance from the reference image.

How does AI reference-based inpainting work?

The API combines visual feature extraction (from the reference image) with context-aware inpainting (using the source image's surroundings). The masked region is regenerated to match the reference's style, texture, color, and content while blending with the surrounding context so the fill reads as part of the original scene.

What inputs does the API need?

Three files inside a zip archive: the source photo to edit, a black-and-white mask defining the area to fill (white = inpaint, black = keep), and a reference image whose content fills the masked region. JPEG, PNG, and WebP supported up to 4096x4096px per file.

How precise does the mask need to be?

Slightly oversized masks (a few extra pixels around the target region) generally produce better blending than pixel-tight masks, since they give the model more context to blend the fill. Hand-drawn brush strokes from a consumer app's eraser tool work for most cases. Very tight or jagged masks can leave edge artifacts.

What kind of reference images work best?

References that share lighting, perspective, and material characteristics with the target region produce the most invisible fills. For fashion: pull a fabric swatch from another product shot. For product photography: use a similar product's surface or an isolated detail shot. The closer the reference matches the scene's conditions, the cleaner the result.

How natural does the inpainted region look?

The fill blends naturally with the surrounding image. Lighting direction, shadow placement, perspective, and color temperature are all matched to the source's context. The reference drives the content, while the source's conditions drive the blending, no visible seams, no obvious style breaks.

What happens if the inpainting fails or the mask is bad?

The API returns a structured response for every call. Successful calls return the inpainted image URL. Failures return an error code with a descriptive message so your integration can handle the case (retry with a refined mask, fall back, or surface to the user). Transient infrastructure issues are retried internally before the call completes.

Do I need rights to the source and reference images?

Yes. Use images you own, have licensed, or have permission to modify. The output is derivative of both inputs, so usage rights follow whichever input has more restrictions. For commercial use, prefer your own photography or properly licensed stock.

How much does the reference inpainting API cost?

$0.55 per image with simple fixed pricing. Sign up and get $10 in free GPU credits to try it out. Contact us for volume pricing on bulk workloads.

Can I process images in bulk?

Yes. Batch endpoints with async webhook delivery. Process thousands of images per job, each with a different mask and reference. Ideal for catalog variation rollouts, fashion pattern swaps, and creative pipelines.

How fast is the reference inpainting 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

On-Model / FashionProduct ImageryAd Creative

Fill in the gaps with reference-guided AI

Reference Based Inpainting API

Inpaint image regions using a reference image as guidance. Match style, texture, and context from your reference with one API call.

Get API access API Docs

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

Try it yourself

Input

image_url *

Your request will cost $0.55

Output

Example output

<5s

Average processing time

$0.55

Per image, fixed pricing

Ref-guided

Style-matched inpainting

Batch

Bulk processing with webhooks

How it works

No pipeline to build, no models to configure.

Send image, mask, and reference

Provide the source image, a mask defining the area to fill, and a reference image to guide the output.

AI inpaints from reference

Runflow's pipeline fills the masked region using the reference image as a style and content guide, blending seamlessly with the surrounding context.

Get your result

Receive the inpainted image with reference-matched content. Typically under 5 seconds.

Inpaint with a reference image as guide

Unlike standard inpainting that guesses what to fill, reference-based inpainting uses a second image to guide the output. The result matches the style, texture, and context of your reference.

Reference image drives the inpainted content

Matches texture, color, and style from the guide

Far more predictable than prompt-only inpainting

Seamless blending with surrounding context

The filled region blends naturally with the rest of the image. Lighting, shadows, and perspective are preserved so the edit is invisible.

Lighting and shadow consistency

Perspective-aware fill

No visible seams or color mismatch

Fashion and product workflows

Swap patterns, fix damaged regions, or extend product shots using a reference. Ideal for catalog consistency, virtual try-on pipelines, and creative asset generation.

Swap fabric patterns using a reference swatch

Fix damaged or occluded regions in product photos

Extend or modify garments with reference guidance

Simple mask + reference input

Send the source image, a mask defining the inpaint region, and a reference image. The API handles the rest: no model training, no prompt engineering.

Source image + mask + reference = result

No prompt engineering required

Works with any standard image format

Reference-based inpainting for every workflow

From fashion catalogs to creative pipelines. Guide your inpainting with a reference instead of guessing.

Fashion & Apparel

Swap fabric patterns, fix garment defects, or extend clothing using reference swatches. Maintain catalog consistency across product variations.

Product Photography

Fix damaged or incomplete product shots by inpainting from a reference. Ensure every product image meets listing standards.

Creative & Advertising

Generate creative variations of campaign imagery using reference-guided fills. Maintain brand consistency while exploring new compositions.

Virtual Try-On Pipelines

Use reference garment images to fill clothing regions on model photos. Build virtual try-on experiences with accurate fabric rendering.

“

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

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

Inpaint with a reference, not a guess. In one API call.

$0.55/image. Reference-guided inpainting that matches style, texture, and context, no prompt engineering. Start with $10 in free GPU credits on signup.

Get API access Browse all solutions

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