Closed cmahnke closed 2 months ago
a little thoughts:
the parameter to the Image.save
function is best to call xmp
(for encoder probably too?)
the parameter type is preferred to be bytes at first stage (different image formats may require different encoding, it will not always be UTF-8) - in future it can be str
or dict
, but bytes
should be supported, imho.
From my reading of https://github.com/python-pillow/Pillow/discussions/8269#discussioncomment-10201110, you're requesting this for both JPEG and MPO?
It's certainly easier to focus on JPEG first, but I wouldn't mind if MPO is also covered. Especially since this would enable creating UltraHDR JPEGs as well. but I suspect this to be a minor use case currently.
Great, thanks! exiftool
reports entries from faked entry using the linked PR. Is there another test you would to be conducted? I don't have Photoshop or similar...
The PR will actually allow XMP data to be written to both JPEG and MPO images.
from PIL import Image
for ext in ('.jpg', '.mpo'):
with Image.open("Tests/images/hopper.png") as im:
im.save('out'+ext, xmp=b"XMP test")
with Image.open('out'+ext) as reloaded:
assert reloaded.info["xmp"] == b"XMP test"
@radarhere: Yes thanks! I could already confirm it. But I wasn't able to add a valid XMP entry to the second image of an MPO. But haven't checked / investigated for a few weeks.
I've pushed a commit to the PR so that it uses the image's info
dictionary as well. That will allow you to do this -
from PIL import Image
im = Image.open("Tests/images/hopper.png")
second_im = Image.new("RGB", (128, 128))
second_im.info["xmp"] = b"Second frame"
im.save("out.mpo", save_all=True, append_images=[second_im])
While the change for this was released in Pillow 11, my plan to allow im.info["xmp"]
to be used for MPO frames has met some resistance - #8479
I've created #8483 with a new plan, using im.encoderinfo
to set XMP data for MPO frames.
This is a feature request: As a followup of this discussion it might be possible to add a simple XMP write support using the provided infrastructure (that is by passing image.encoderinfo["extra"]). This approach doesn't account for large XMP Data (>64Kb), but would be an improvement.
For an discussion on XMP, especially large XMP fields, see #5076, there are also sample images linked.
To keep the interface simple, it should only accept
str
in the first iteration, later on some XML structure, requirements for adict
interface would need full support for the XML data model like XML namespace support and lists of elements with the same name.See also #8069