fugo

Imaging

Default quality, resampling filter, anchor, background colour and EXIF handling for image processing, and the metadata of published images.

config.toml
[imaging]
  quality = 75
  resampleFilter = "box"
  anchor = "smart"
  bgColor = "#ffffff"
  hint = "photo"
  stripMetadata = true
  copyright = ""
  author = ""
  titleFromFileName = false
  [imaging.exif]
    includeFields = ""
    excludeFields = ""
    disableDate = false
    disableLatLong = false
quality
JPEG and WebP quality, 1–100 (75).
resampleFilter
The filter for resizing: box (default), lanczos, catmullRom, mitchellNetravali, linear, nearestNeighbor and others; see Image processing.
anchor
Where fill and crop keep the image: smart (default), center, topLeft, …
bgColor
The background for images converted to a format without transparency (JPEG).
hint
The WebP encoding hint: photo (default), picture, drawing, icon, text.
stripMetadata
Publish JPEG, PNG and WebP images without their metadata (true by default; fugo’s own setting). See Image metadata.
copyright, author, titleFromFileName
A copyright notice ({year} in it is the build’s year), an author and the file name as the title, written into every JPEG, PNG and WebP image fugo publishes, unless a page resource gives its own copyright and author (fugo’s own settings; they need stripMetadata). See Copyright notice.
exif.includeFields, exif.excludeFields
Regular expressions of the EXIF tags to keep in img | exif. By default fugo leaves out large and technical tags.
exif.disableDate, exif.disableLatLong
Do not read the date or the location.

Processed images are cached in resources/_gen/images; keep that directory between builds (or commit it) to avoid processing images again. See Caching.

Image metadata#

A photo carries metadata: the camera, when it was taken and, from a phone, where (its GPS position). Processed images (resize, fill, filters, …) never do. The images fugo publishes as they are do: page resources, assets whose URL a page prints, and the files of static/.

By default, fugo removes from those JPEG, PNG and WebP files EXIF (camera, dates, GPS position, thumbnail), XMP, IPTC, comments, PNG text chunks and other vendor data, and anything appended after the image (such as the video of a motion photo). What changes how the image looks stays: the colour profile and other colour information, and the EXIF orientation, which a photo taken sideways needs to show upright.

Only the published copies change: your files keep their metadata, and img | exif still reads it. Other formats (GIF, TIFF, BMP, AVIF, HEIC, …) are published as they are. Copies published by an earlier build, and your Git history if the images are committed, keep their metadata.

To publish your images byte for byte as they are, as the Go implementation does, set stripMetadata = false.

fugo can write a copyright notice, an author and a title into the JPEG, PNG and WebP images it publishes, the processed ones included:

config.toml
[imaging]
  copyright = "© 2019–{year} by Jane Doe, all rights reserved"
  author = "Jane Doe"
  titleFromFileName = true

They go into the image’s EXIF (Copyright and Artist, which photo applications show) and XMP (dc:rights, dc:creator and dc:title, which Google Images shows as the copyright notice and the creator), in place of the image’s own metadata: they need stripMetadata. With titleFromFileName, harbour_view.jpg gets the title harbour_view, and an image processed from it the same. Each setting holds at most 1000 bytes, without line breaks.

{year} in copyright is the year of the build: of --clock when it is given, else of the system clock, in the default language’s time zone (timeZone), the year now() gives its templates. Built in 2027, the setting above writes © 2019–2027 by Jane Doe, all rights reserved. It is the only placeholder: any other brace in copyright, such as a misspelt {yaer}, is an error, so that a typo does not go into your images. The notice is written when an image is published, after the image cache, so a new year does not process the images again. The author is written as it is.

A photo by someone else, such as one from Wikimedia Commons under a Creative Commons licence, needs its own notice. A page resource’s copyright and author params (metadata) are written into it, and into every image processed from it, in place of [imaging]’s:

content/posts/harbour/index.md
+++
title = "The harbour"
[[resources]]
  src = "harbour.jpg"
  [resources.params]
    author = "Ada Example"
    copyright = "© Ada Example, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)"
+++

They replace both of [imaging]’s, so that the photo never carries your copyright next to its author: one the resource does not set is not written. The title is still the file name (titleFromFileName). Their values follow the rules of [imaging]’s: {year} is the build’s year, and each holds at most 1000 bytes of text without line breaks (in YAML, wrap a long one with >-: > keeps a final line break); a value that breaks them, or is not text, stops the build with an error naming the page and the resource. Like [imaging]’s, the notice is written when the image is published: changing it processes no image again. Other params, such as a license your templates print, are yours. With stripMetadata = false, images are published as they are and these params do nothing.

A photo several pages publish carries one notice: that of the first of them, in language order (the default content language first, then by weight), whose metadata gives one. A page without the metadata takes nothing away. The translations of a bundle share its files, or each publishes a copy of them (on a site with several hosts, with [markup.goldmark] duplicateResourceFiles = true, or for a content file that is not Markdown): every copy carries that notice. So do the resources that content adapters add for each language from one image.

    Type to search titles, headings and text.