hi, jcupitt. sorry to trouble you again
i try below codes to make some pixels be transparent. the result is fine. but i guess some meta data were lost (photoshop open it and look correct, but my uv-printer can't recognize the spot color channel)
`
def erase_outside_mask():

input_path = "cat.tif"
output_path = "result.tif"
mask_path = "mask.png"
# read tiff
image = pyvips.Image.new_from_file(input_path, access="sequential")
# read mask png
mask = pyvips.Image.new_from_file(mask_path, access="sequential")
# extract alpha band
mask_alpha = mask.extract_band(3)
# erase image
image = mask_alpha.ifthenelse(image, [255, 255, 255, 255, 0, 255])
# save image
image.tiffsave(output_path, compression="deflate")

`
the input tiff image size is 7MB and include "Adobe Photoshop Document Data Block" meta data. but the output tiff image size is 2.5M without that meta data. I don't know why. can you help me?
tiffinfo the input file:

tiffinfo the output file:

all the resources in this zip file
pyvips-test.zip

You're right, the libvips TIFF read and writer will carry over icc, xmp, iptc and photoshop data, but they won't read and write "photoshop document data block", whatever that is :(

I'll have a look.

it seem pyvips keep the spot color channel, but lost the data belong to it.
event i only read then write the tiff file, I found the spot color channel data was missing. but i open the result tiff file with photoshop and just save it as a copy. then the spot color channel data come back.

   input_path = "cat.tif"
   output_path = "demo.tif"
   # read source tiff
   image = pyvips.Image.new_from_file(input_path, access="sequential")
   # save image
   image.tiffsave(output_path,compression='lzw', xres=image.xres, yres=image.yres,subifd=True)

It looks difficult to support:

https://www.amyspark.me/blog/posts/2021/11/14/reading-and-writing-tiff-psds.html

It seems photoshop stores extra image data in a private TIFF tag, and it is not well documented.

libvips could just copy the data block over, maybe that would be enough?

how to copy then? I have no any idea if it is enough or not.

I made this prog:

#!/usr/bin/env python

import pyvips

input_path = "cat.tif"
output_path = "result.tif"
mask_path = "mask.png"
# read tiff
image = pyvips.Image.new_from_file(input_path, access="sequential")
# read mask png
mask = pyvips.Image.new_from_file(mask_path, access="sequential")
# extract alpha band
mask_alpha = mask[3]
# erase image
image = mask_alpha.ifthenelse(image, [255, 255, 255, 255, 0, 255])
# save image
image.tiffsave(output_path, compression="deflate")

And with this patched libvips it made this image:

result.zip

Does that work on your printer?

no, the RIP software parse it with wrong size and multiple image。

Ah that's a shame. Then I don't think libvips will be able to work with these files, they are too non-standard.

got it. maybe i should change my solution then. anyway,thanks for you help so much