pictomancer

n8n community node for Pictomancer.ai - image optimization (resize, compress, convert, crop, pipeline)

Package Information

Released: 2 mo ago
Downloads: 0 weekly / 0 monthly
Latest Version: 0.6.0

Available Nodes

Documentation generating…

Documentation

n8n-nodes-pictomancer

n8n community node for Pictomancer.ai - image optimization in your workflows: resize, compress, convert (JPEG/PNG/WebP/TIFF/GIF/AVIF), crop, and multi-step pipelines.

Install

In n8n: Settings → Community Nodes → Install and enter n8n-nodes-pictomancer.

Or with npm (self-hosted):

npm install n8n-nodes-pictomancer

Credentials

Create a Pictomancer API credential:

  • API Key - Bearer token from app.pictomancer.ai. Optional: without it you are on the free tier (50 requests per IP).
  • Base URL - defaults to https://api.pictomancer.ai.

Operations

Operation Input Output
Analyze URL or binary JSON metadata (size_bytes)
Resize URL or binary Binary image
Compress URL or binary Binary image
Convert URL or binary Binary image
Crop URL or binary Binary image
Pipeline URL or binary Binary image
  • Input Type switches between a source URL and binary data from the previous node (sent as a data: URI).
  • Image operations output binary data (default property data) plus size_bytes, mime_type, and billed in the item JSON. billed is false when compress produced no size gain - those requests are free.
  • Options exposes quality (q), output format, metadata strip, lossless, and the AVIF effort knob.
  • Quality Target (SSIM) (compress and convert): ask for the smallest file with SSIM >= target (0-1) instead of picking a q value. Mutually exclusive with Quality (q); jpeg, webp and avif only; compress needs an explicit Output Format. When the search runs, the item JSON also carries quality_target, quality_achieved, quality_final_q, and quality_encodes.
  • Crop Mode (crop) switches between three mutually exclusive modes: Manual (X/Y/Width/Height), Smart (Gravity picks the window; needs Width/Height), and Trim (removes a uniform background border; Threshold defaults to 10). When a crop actually trims, the response carries X-Pictomancer-Trim-Left/-Top/-Width/-Height headers.
  • Fill Width / Fill Height / Gravity (resize, under Options): set both to resize and smart-crop to exact dimensions in one call instead of Scale/Scale X/Scale Y.
  • Autorot (resize, compress, convert, crop, under Options): apply EXIF orientation before processing.
  • Denoise (resize, compress, convert, crop, under Options): median denoise before the operation, radius 1-3 (window 3x3 to 7x7). Base price.
  • Equalize (resize, compress, convert, crop, under Options): auto-contrast (value-channel histogram equalisation, hue and saturation preserved) before the operation.
  • Sharpen (resize, compress, convert, crop, under Options): unsharp-mask sharpen after the operation.
  • Optimize AI-Generated: one-call web optimization of generator output (webp/avif/jpeg/png, max dimension, quality target)
  • Pipeline takes a JSON list of steps, e.g. [{ "type": "resize", "params": { "scale": "0.5" } }, { "type": "convert", "params": { "format": "webp" } }].

Development

npm install
npm test
npm run build

License

MIT

Discussion