Return a 0..1 fraction for x/y/w/h given the deck's unit system.

Return a 0..1 fraction for x/y/w/h given the deck's unit system.

Return text size in points. Preferred Image2PPTX field is size_ratio/size_pct: source text height as a fraction/percent of the source image h

Write a labeled visual index of sliced icons for mandatory QA.

Yield (start, end) index ranges of True runs, merging gaps shorter than min_gap.

Set text color alpha for lightly visible GPT-vision text overlays.

Set latin + east-asian + complex-script typeface so CJK renders correctly.

Set latin + east-asian + complex-script typeface so CJK renders correctly.

Best-effort soft drop shadow for card-like shapes.

Best-effort soft drop shadow for card-like shapes.

Return (l, t, r, b) bounding box of pixels with alpha > min_alpha, or None.

Tune defaults for frame layers vs icon sheets without hiding explicit flags.

Projection-based segmentation: split content into rows by horizontal gaps, then each row into items by vertical gaps. Robust when the AI grid is i

Connected-component segmentation by alpha transparency. Returns list of (bbox, index) sorted top-to-bottom, left-to-right. Components are bas

Resize source to preview size using the preview aspect as authority.

Approximate polygon vertices for preview rendering of non-rect shapes.

Approximate polygon vertices for preview rendering of non-rect shapes.

Fill pinholes/gaps without reshaping existing anti-aliased edges. A direct max/min filter on the grayscale alpha channel makes straight lines

Median color of the image border / corners (the flat key background).

Add an <a:alpha> to a shape's solid fill so cards can be translucent.

Add an <a:alpha> to a shape's solid fill so cards can be translucent.

Channels that carry the key color's energy (e.g. [1] for green, [0,2] for magenta).