Clipdrop

Edit and generate images with Clipdrop by Stability AI. Agents remove and replace backgrounds, clean up and upscale photos, erase text, create product photography, and generate images from text or sketches, saving every result to FlowRunner file storage.

11 actions API key available

Clipdrop website ↗ Platform Documentation ↗ Capability data verified 2026-07-27

Raw product shots arrive from the supplier in the shared drive

Remove Background cuts each product to a transparent PNG, preserving hair-fine edges, saved to FlowRunner file storage

Replace Background composites the hero SKUs into the campaign scene described in the brief's prompt

Upscale Image raises the approved compositions to the exact banner resolution the site needs

Each composite is compared against the product's real attributes, because a generated scene can imply props, materials, or scale the product does not have

The brand owner approves each hero composite before anything reaches the storefront

Cleanup erases the stray artifacts the reviewer masks, white pixels marking what dies

Approved assets post to the campaign channel with durable URLs and their credit usage logged

What This Integration Enables

Clipdrop packages Stability AI's editing and generation models as single-purpose calls, and this connector turns them into pipeline steps with the storage already handled: every result saves to FlowRunner file storage and returns a durable URL plus its credit usage, so downstream steps and cost tracking both just work. FlowRunner agents remove and replace backgrounds, turn plain shots into studio-grade product photography, erase objects with masks and text without them, upscale to exact pixel targets, generate variations of an existing composition, create images from prompts or sketches, and compute depth and surface-normal maps from portraits for relighting and 3D work. In practice, the leverage shows up as: - One master cutout feeding every campaign scene, size, and season without a reshoot - Spend that stays auditable, because every call returns its credit usage alongside the durable URL - Format matrices filled by upscaling and variation instead of designer queue time - Generated composites that reach the storefront only through a brand approval The stance this page takes: generation is cheap, judgment is not. When a background is invented rather than photographed, the question shifts from "is this image good" to "is this image true enough to sell with", and that question belongs to a person. The pipeline runs at machine speed up to that [human-in-the-loop](/concepts/human-in-the-loop) gate, and after approval, at machine speed again.

Without FlowRunner

Hero images wait on a designer Every campaign scene needs studio time or an editing queue

One photo, one use The supplier shot works for the listing and nothing else, so banners get improvised

Generated content ships unexamined AI composites go live with whatever props and scale the model dreamed up

With FlowRunner

Scenes generated per campaign The same cutout lands on a marble counter for spring and a linen table for fall

Every asset in every size Upscaling and variation generation turn one photo into the full format matrix

Generation gated by brand judgment A person approves what the model composed before a customer sees it

Use Case Scenarios

One supplier photo, a season of assets

A single product shot enters the flow. Remove Background produces the master cutout, Replace Background stages it into the spring campaign scene, Reimagine generates three alternate takes for A/B testing, and Upscale Image sizes the winner for homepage, email, and social formats. The results archive to [S3](/integrations/s3) with durable URLs, and the listing that used to have one image has a coordinated set.

The inherited asset library, rescued

Marketing takes over a brand with years of inconsistent imagery: watermarked exports, busy backgrounds, low-resolution favorites. The agent batches the library through Remove Text for the watermarks, Cleanup with masks for the clutter that inpainting should erase, and Upscale Image for the print candidates, logging each file's credit usage as it goes. The keepers come out clean and durable-linked; the unrescuable get flagged for reshoot. A quarter's worth of retouching requests becomes a supervised batch.

Concept to comp before the meeting ends

A campaign brainstorm produces rough sketches and a paragraph of art direction. Sketch to Image turns the drawings into rendered concepts guided by the brief's language, Text to Image explores directions nobody sketched, and Portrait Depth Estimation prepares the chosen hero portrait for the depth-of-field treatment the designer wants. The creative review starts from rendered options instead of descriptions, the strongest direction gets refined with Reimagine variations, and the deck ships the same day.

Human-in-Loop Highlight

Replace Background and Product Photography compose the risky pixel: they put a real, sellable product into a scene that never existed, and a customer will reasonably read that scene as information. A generated kitchen counter that implies the wrong size, invented props that suggest included accessories, a surface texture that flatters the material out of recognition: each is a return, a complaint, or a marketplace policy problem wearing a beautiful image. So in FlowRunner, generated composites are drafts by definition. The agent produces them, stages them with durable URLs, and routes each to the brand owner with the original photo beside it. Approval is what promotes a composite from file storage to storefront. Nothing the model invents reaches a customer without a person having compared it to the product it claims to show.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via preferred channel

Human decides

Agent resumes with decision

Agent Capabilities

11 actions

Background

Remove Background Cuts the subject to a transparent PNG, preserving fine detail like hair and edges, from images up to 25 megapixels.
Replace Background Composites the subject into a new scene generated from a text prompt. The campaign-scene generator this page gates.
Product Photography Builds a clean, well-lit studio scene around a plain product shot. Best fed a simple or transparent background.

Retouching

Cleanup Erases objects, people, or defects using a same-resolution mask where white pixels mark what to remove.
Remove Text Detects and inpaints away all text automatically, no mask required. The watermark and caption eraser.
Upscale Image Super-resolves to an exact target width and height up to 4096 pixels per side, sharpening as it enlarges.

Generation

Reimagine Generates a variation of an input image, keeping composition and subject while reworking details and style. No prompt needed.
Text to Image Creates a new 1024x1024 image from a prompt describing subject, style, lighting, and composition.
Sketch to Image Turns a rough sketch into a detailed image guided by a text prompt. The concept-to-comp step.

Portrait Analysis

Portrait Depth Estimation Produces a grayscale depth map of a portrait for depth-of-field, compositing, and relighting work.
Portrait Surface Normals Computes a surface-normal map encoding per-pixel orientation, for relighting and 3D reconstruction.

Frequently Asked Questions

What can FlowRunner do with Clipdrop?

FlowRunner agents can run Remove Background, Replace Background, and Product Photography in Clipdrop, plus 8 more actions.

Does connecting Clipdrop to FlowRunner require OAuth?

No. Clipdrop connects to FlowRunner with an API key, no OAuth flow required.

Can Clipdrop trigger a FlowRunner workflow automatically?

Clipdrop doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with Clipdrop

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us