n8n Edit Image node: resize, crop, and watermark your images without a third-party service

A supplier hands you their catalog as 4000×3000 photos, and your store expects square 1200×1200 thumbnails with a logo watermark. Doing this by hand in Photoshop is fine for a handful of images, but it quickly becomes unsustainable on a catalog with hundreds of SKUs — and sending them to a SaaS image-processing service means paying per transformation and routing your photos through a third party. n8n ships a native solution for exactly this case: the Edit Image node, which resizes, crops, watermarks, and converts your images directly on your own instance, with no external API call.

An unusual node: local processing, not an API call

Unlike almost every other n8n node, Edit Image doesn't talk to any remote service. It hands the work to GraphicsMagick, an image-processing library installed on the machine running n8n. In practice:

• Official n8n Docker image (self-hosted, see our Docker installation guide): GraphicsMagick ships preinstalled, nothing to do.
• Self-hosted without Docker (manual VPS setup): you need to install it yourself, for example sudo apt install graphicsmagick on a Debian/Ubuntu distribution.
• Queue mode with workers (see our queue mode and Redis guide): GraphicsMagick must be present on every worker container, not just the main instance — a common oversight that produces a "GraphicsMagick module not found" error on workers even though the node works fine locally.
• n8n Cloud: the node depends on a system binary rather than an API, which makes its behavior more sensitive to the managed environment. Test the exact operation you need on a small sample before building a critical pipeline around it.

This local processing has a direct benefit for sensitive data: images (photos of people, scanned documents, attachments) never leave your infrastructure to be resized, unlike a call to a third-party image-processing API.

Available operations

The node exposes thirteen operations, combinable within a single instance via Multi Step mode:

Operation	Purpose	Key parameters
Resize	Resize	width, height, mode (Ignore Aspect Ratio, Maximum Area, Minimum Area, Only if Larger, Only if Smaller, Percent)
Crop	Crop	width, height, position X/Y
Text	Add text (watermark)	text, font size, color, position X/Y
Composite	Overlay an image (logo)	binary property of the image to overlay, operator, position X/Y
Rotate	Rotate	angle (-360 to 360°), background color
Border	Add a border	width, height, color
Blur	Blur	intensity, sigma
Draw	Draw a shape	circle, line, rectangle, color
Shear	Shear the image	X/Y degrees
Transparent	Make a color transparent	target color
Create	Create a blank image	width, height, background color
Get Information	Read metadata	none (returns resolution, format…)
Multi Step	Chain several operations	ordered list of the operations above

Three global settings apply regardless of the operation: Data Property Name (the input binary field, data by default), output Format (bmp, gif, jpeg, png, tiff, webp), and Quality (0-100, for jpeg/png/tiff) — that last setting is what lets you shrink a batch's file size without reaching for an external compression tool.

Cropping isn't cosmetic: what the research says

Resizing and cropping a product photo isn't just about aesthetics — it's one of the better-documented levers in e-commerce conversion. A study by Li, Wang, and Chen, presented at the Pacific Asia Conference on Information Systems (PACIS) in 2014 and based on image analysis across a large volume of real product photos, found that photos where the product fills a larger share of the frame, against a less cluttered background, generate stronger purchase intent than photos where the product is small and lost in its surroundings (Li, Wang, Chen 2014, PACIS — see on Google Scholar). A tight, centered crop via the Crop operation, rather than a plain resize that keeps the entire original frame, moves in the direction the research identifies as favorable to conversion — a setting that takes thirty seconds to automate once the pipeline is in place, across hundreds of photos.

Example: supplier photos → watermarked catalog thumbnail pipeline

A typical setup to standardize a batch of supplier photos before publishing:

Trigger (new file in Drive/S3) → Read binary → Edit Image (Multi Step) → Write to the platform

Inside a single Edit Image node, in Multi Step mode, three operations are stacked in order:

1. Resize, Maximum Area mode, 1600×1600 — brings any oversized image down to a reasonable weight without ever upscaling it or distorting its ratio.
2. Crop, 1200×1200, position computed to center the subject — produces the square thumbnail most e-commerce platforms expect.
3. Text, a subtle watermark in the bottom-right corner (brand name, low opacity, font size proportional to the image).

Then, in the node's global options: Format set to webp and Quality at 82 — a solid size/quality tradeoff for the web, plenty for catalog display while noticeably cutting bandwidth compared to an uncompressed JPEG. The resulting file then goes to Shopify or WooCommerce via the platform's native node (see our Shopify and WooCommerce guides), or to an S3 bucket for archiving — the exact pattern is covered in our guide to archiving files to S3. If the end goal is publishing a full product listing rather than a single image, this pipeline slots naturally upstream of the AI product-listing generation pipeline: standardized image on one side, generated copy on the other, published together at the end.

Text watermark or logo: Text vs. Composite

Two approaches depending on the need: the Text operation stamps a string of characters directly onto the image — quick to configure, plenty for a brand name or a copyright line, but limited to typography. The Composite operation overlays a second binary image (a PNG logo with a transparent background, for instance) at a given position, with a blend operator (Over, Multiply, Atop…) — the method to reach for once the watermark is a logo rather than text, or needs to follow a precise brand guideline. Both images (source and logo) need to be present as binary data on the item by the time the Composite node runs: a Merge node upstream, with two inputs, brings them together into the same item before processing.

Anonymization via blur: a use case of its own

The Blur operation applied to a region cropped beforehand (via a targeted Crop on a face or license plate detected by a vision node — see our guide to image analysis with AI) lets you build a visual anonymization pipeline before archiving or publishing — a visual counterpart to the text pseudonymization techniques covered in our GDPR guide to anonymizing personal data. Because processing stays local (GraphicsMagick, no third-party API), no sensitive image passes through an external service during the operation.

Processing a batch without exhausting memory

GraphicsMagick loads one image into memory at a time, not the whole batch — the real risk on a catalog with hundreds of files comes from binary data piling up within the n8n execution itself if nothing purges it between items. A Loop Over Items node that processes the batch in chunks of 20 to 50 images, rather than in a single pass, keeps the instance's memory usage under control; the finer-grained settings (the NODE_OPTIONS variable, binary execution pruning) are detailed in our guide to large files in n8n.

When to look elsewhere

The Edit Image node covers the bulk of everyday raster processing, but it stays within GraphicsMagick's scope: no AI-driven retouching (automatic background removal, generative upscaling), no vector formats. For those needs, the Edit Image+ community node adds extra effects — enable it as a community node on a self-hosted instance, subject to checking the activation policy that applies to your n8n Cloud plan. For background removal or model-driven quality enhancement, an HTTP Request call to a specialized API, followed by an Edit Image node for the final crop and format, is usually still the most reliable setup.

Summary

The Edit Image node processes your images directly on your own n8n instance, with no API call and no per-transformation cost: resizing, cropping, text or logo watermarking, and format conversion all combine into a single node instance via Multi Step mode. For a batch of product photos, the sequence Resize → Crop → Text (or Composite) → Format/Quality standardizes an entire catalog in one reusable workflow. If this image pipeline feeds into full product listings, the AI product-listing generation guide picks up on the copy side; and if those images arrive as email attachments that need triaging before archiving, the Inbox AI Pack provides the triage building block that can precede this image processing in the same workflow.