twat-image 2.7.7

A Python utility for converting images to include alpha channels based on grayscale values, and other image manipulations.

twat-image

twat-image is the image domain package for the twat plugin ecosystem. It owns deterministic image operations and delegates AI generation or editing to twat_genai when that optional package is installed.

Install

pip install twat-image

For development:

pip install -e ".[test,dev]"

Python API

from PIL import Image
from twat_image import igray2alpha, scale_image, read_image_metadata

image = Image.open("logo.png")
mask = igray2alpha(image, color="black", white_point=0.9, black_point=0.1)
mask.save("logo-alpha.png")

small = scale_image(image, width=512)
small.save("logo-512.png")

print(read_image_metadata("logo.png"))

Deterministic helpers include:

• igray2alpha / gray2alpha: grayscale-derived alpha masks.
• alpha_from_diff: alpha from the grayscale difference of two images.
• normalize_image: auto-contrast and optional equalization.
• scale_image, crop_image, outcrop_image: basic geometry.
• convert_image: format conversion through Pillow.
• read_image_metadata, find_duplicate_images: dimensions, format, and average-hash duplicate grouping.

CLI

python -m twat_image --help
python -m twat_image gray2alpha input.jpg output.png --color black
python -m twat_image scale input.png output.png --width 512
python -m twat_image crop input.png crop.png 10 10 300 300
python -m twat_image outcrop input.png padded.png --left 40 --right 40 --top 20 --bottom 20
python -m twat_image normalize input.png normalized.png
python -m twat_image metadata input.png

The legacy imagealpha script remains as a compatibility alias for the original grayscale-to-alpha command.

AI image boundary

twat-image does not contain provider clients. Use generate_image() and edit_image() as thin adapters to twat_genai:

from twat_image import generate_image, edit_image

generate_image("a black cat drawn as clean vector art", output_dir="out")
edit_image("make it look like ink on paper", "input.png", output_dir="out")

Install and configure twat-genai separately for OpenAI, Gemini/Nano Banana, Fal, or future provider engines.

Reference scripts

Reusable deterministic work from reference/bin-img-vid/imggray2alpha, imgalphafromdiff.py, imgscale.py, imgcrop, imgoutcrop, and imgdedup is represented by the APIs above. Provider-heavy scripts such as imgquotio.py and imgnanobanatrans.py belong in twat_genai; this package only calls that boundary.

Local shell workflow glue remains in reference/bin-img-vid/ because it encodes workstation-specific pipelines rather than reusable library behavior.