Generating renditions in Python

Rendered versions of original images generated by the Wagtail {% image %} template tag are called “renditions”, and are stored as new image files in the site’s [media]/images directory on the first invocation.

Image renditions can also be generated dynamically from Python via the native get_rendition() method, for example:

newimage = myimage.get_rendition('fill-300x150|jpegquality-60')

If myimage had a filename of foo.jpg, a new rendition of the image file called foo.fill-300x150.jpegquality-60.jpg would be generated and saved into the site’s [media]/images directory. Argument options are identical to the {% image %} template tag’s filter spec, and should be separated with |.

The generated Rendition object will have properties specific to that version of the image, such as url, width and height. Hence, something like this could be used in an API generator, for example:

url = myimage.get_rendition('fill-300x186|jpegquality-60').url

Properties belonging to the original image from which the generated Rendition was created, such as title, can be accessed through the Rendition’s image property:

    >>> newimage.image.title
    'Blue Sky'
    >>> newimage.image.is_landscape()
    True

See also: How to use images in templates

Generating multiple renditions for an image

You can generate multiple renditions of the same image from Python using the native get_renditions() method. It will accept any number of ‘specification’ strings or Filter instances, and will generate a set of matching renditions much more efficiently than generating each one individually. For example:

image.get_renditions('width-600', 'height-400', 'fill-300x186|jpegquality-60')

The return value is a dictionary of renditions keyed by the specifications that were provided to the method. The return value from the above example would look something like this:

{
    "width-600": <Rendition: Rendition object (7)>,
    "height-400": <Rendition: Rendition object (8)>,
    "fill-300x186|jpegquality-60": <Rendition: Rendition object (9)>,
}

Caching image renditions

Wagtail will cache image rendition lookups, which can improve the performance of pages which include many images.

By default, Wagtail will try to use the cache called “renditions”. If no such cache exists, it will fall back to using the default cache.

Prefetching image renditions

When using a queryset to render a list of images or objects with images, you can prefetch the renditions needed with a single additional query. For long lists of items, or where multiple renditions are used for each item, this can provide a significant boost to performance.

Image QuerySets

When working with an Image QuerySet, you can make use of Wagtail’s built-in prefetch_renditions queryset method to prefetch the renditions needed.

For example, say you were rendering a list of all the images uploaded by a user:

def get_images_uploaded_by_user(user):
    return ImageModel.objects.filter(uploaded_by_user=user)

The above can be modified slightly to prefetch the renditions of the images returned:

def get_images_uploaded_by_user(user):
    return ImageModel.objects.filter(uploaded_by_user=user).prefetch_renditions()

The above will prefetch all renditions even if we may not need them.

If images in your project tend to have very large numbers of renditions, and you know in advance the ones you need, you might want to consider specifying a set of filters to the prefetch_renditions method and only select the renditions you need for rendering. For example:

def get_images_uploaded_by_user(user):
    # Only specify the renditions required for rendering
    return ImageModel.objects.filter(uploaded_by_user=user).prefetch_renditions(
        "fill-700x586", "min-600x400", "max-940x680"
    )

Non Image Querysets

If you’re working with a non Image Model, you can make use of Django’s built-in prefetch_related() queryset method to prefetch renditions.

For example, say you were rendering a list of events (with thumbnail images for each). Your code might look something like this:

def get_events():
    return EventPage.objects.live().select_related("listing_image")

The above can be modified slightly to prefetch the renditions for listing images:

def get_events():
    return EventPage.objects.live().select_related("listing_image").prefetch_related("listing_image__renditions")

If you know in advance the renditions you’ll need, you can filter the renditions queryset to use:

from django.db.models import Prefetch
from wagtail.images import get_image_model


def get_events():
    Image = get_image_model()
    filters = ["fill-300x186", "fill-600x400", "fill-940x680"]

    # `Prefetch` is used to fetch only the required renditions
    prefetch_images_and_renditions = Prefetch(
        "listing_image",
        queryset=Image.objects.prefetch_renditions(*filters)
    )
    return EventPage.objects.live().prefetch_related(prefetch_images_and_renditions)

Model methods involved in rendition generation

The following AbstractImage model methods are involved in finding and generating renditions. If using a custom image model, you can customize the behavior of either of these methods by overriding them on your model:

class wagtail.images.models.AbstractImage
get_rendition(filter: Filter | str) AbstractRendition

Returns a Rendition instance with a file field value (an image) reflecting the supplied filter value and focal point values from this object.

Note: If using custom image models, an instance of the custom rendition model will be returned.

find_existing_rendition(filter: Filter) AbstractRendition

Returns an existing Rendition instance with a file field value (an image) reflecting the supplied filter value and focal point values from this object.

If no such rendition exists, a DoesNotExist error is raised for the relevant model.

Note: If using custom image models, an instance of the custom rendition model will be returned.

create_rendition(filter: Filter) AbstractRendition

Creates and returns a Rendition instance with a file field value (an image) reflecting the supplied filter value and focal point values from this object.

This method is usually called by Image.get_rendition(), after first checking that a suitable rendition does not already exist.

Note: If using custom image models, an instance of the custom rendition model will be returned.

get_renditions(*filters: Filter | str) Dict[str, AbstractRendition]

Returns a dict of Rendition instances with image files reflecting the supplied filters, keyed by filter spec patterns.

Note: If using custom image models, instances of the custom rendition model will be returned.

find_existing_renditions(*filters: Filter) Dict[Filter, AbstractRendition]

Returns a dictionary of existing Rendition instances with file values (images) reflecting the supplied filters and the focal point values from this object.

Filters for which an existing rendition cannot be found are ommitted from the return value. If none of the requested renditions have been created before, the return value will be an empty dict.

create_renditions(*filters: Filter) Dict[Filter, AbstractRendition]

Creates multiple Rendition instances with image files reflecting the supplied filters, and returns them as a dict keyed by the relevant Filter instance. Where suitable renditions already exist in the database, they will be returned instead, so as not to create duplicates.

This method is usually called by Image.get_renditions(), after first checking that a suitable rendition does not already exist.

Note: If using custom image models, an instance of the custom rendition model will be returned.

generate_rendition_file(filter: Filter, *, source: File = None) File

Generates an in-memory image matching the supplied filter value and focal point value from this object, wraps it in a File object with a suitable filename, and returns it. The return value is used as the file field value for rendition objects saved by AbstractImage.create_rendition().

If the contents of self.file has already been read into memory, the source keyword can be used to provide a reference to the in-memory File, bypassing the need to reload the image contents from storage.

NOTE: The responsibility of generating the new image from the original falls to the supplied filter object. If you want to do anything custom with rendition images (for example, to preserve metadata from the original image), you might want to consider swapping out filter for an instance of a custom Filter subclass of your design.