Skip to content

GatsbyFitz/Remotion-Short-Creator / .claude/skills/overlay-coordinate-grid

overlay-coordinate-grid

Verified

Place and verify every overlay by reading exact coordinates off a labelled coordinate grid instead of estimating positions. Grid the real footage frame when graphics must line up with something in shot, and grid the overlay's own renders to measure where things actually landed — text bounds, collisions, margins, centring. Use for every overlay composition, whether or not it's tied to footage.

FileKindSize
SKILL.mdskill md7.6 KB
scripts/composite.pyscript3.4 KB
scripts/grid.pyscript4.9 KB

scripts/grid.py

View on GitHub
#!/usr/bin/env python3
"""Draw a labelled coordinate grid over an image so positions can be read off it
exactly rather than estimated.

Works on a footage frame (a screenshot, or a frame pulled from a clip) or on a
render of the overlay itself. The image is normalised to the composition size
first, so every label is in composition pixels and a number read off the sheet
goes straight into code.

usage:
  grid.py IMAGE OUT.png [--size 1920x1080] [--step 100] [--background dark]
          [--region NAME=x0,y0,x1,y1[,step[,zoom]] ...] [--save-normalised PATH]

With no --region, the whole image is gridded at --step. Each --region adds a
zoomed, finer-gridded crop instead; crops are stacked into OUT.png. Transparent
renders are flattened onto --background (dark, light, or #rrggbb) so both light
and dark elements stay visible.
"""
import argparse
from PIL import Image, ImageDraw, ImageFont

BACKGROUNDS = {"dark": (24, 26, 30), "light": (208, 210, 214)}


def label_font(size):
    try:
        return ImageFont.load_default(size=size)
    except TypeError:
        try:
            return ImageFont.truetype("/System/Library/Fonts/Supplemental/Arial.ttf", size)
        except OSError:
            return ImageFont.load_default()


def gridded(image, box, step, scale, name="", font_size=16):
    """Crop `box` from `image`, scale it, and grid it with labels in the image's own
    (unscaled) coordinates."""
    x0, y0, x1, y1 = box
    crop = image.crop(box).resize((round((x1 - x0) * scale), round((y1 - y0) * scale)), Image.LANCZOS)
    draw = ImageDraw.Draw(crop)
    font = label_font(font_size)

    def label(xy, text, colour):
        left, top, right, bottom = draw.textbbox(xy, text, font=font)
        draw.rectangle((left - 2, top - 1, right + 2, bottom + 1), fill=(0, 0, 0))
        draw.text(xy, text, fill=colour, font=font)

    for x in range(-(-x0 // step) * step, x1 + 1, step):
        px = (x - x0) * scale
        draw.line([(px, 0), (px, crop.height)], fill=(0, 255, 255), width=1)
        label((px + 3, 3), str(x), (0, 255, 255))
    for y in range(-(-y0 // step) * step, y1 + 1, step):
        py = (y - y0) * scale
        draw.line([(0, py), (crop.width, py)], fill=(255, 0, 255), width=1)
        label((3, py + 3), str(y), (255, 0, 255))
    if name:
        label((6, crop.height - font_size - 6), name, (255, 255, 0))
    return crop


def parse_region(text):
    name, _, spec = text.partition("=")
    parts = [int(v) for v in spec.split(",")]
    if len(parts) < 4:
        raise SystemExit(f"region {text!r}: need NAME=x0,y0,x1,y1[,step[,zoom]]")
    return name, tuple(parts[:4]), (parts[4] if len(parts) > 4 else 50), (parts[5] if len(parts) > 5 else 1)


def parse_background(value):
    if value in BACKGROUNDS:
        return BACKGROUNDS[value]
    value = value.lstrip("#")
    return tuple(int(value[i:i + 2], 16) for i in (0, 2, 4))


def main():
    parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
    parser.add_argument("image")
    parser.add_argument("out")
    parser.add_argument("--size", default="1920x1080")
    parser.add_argument("--step", type=int, default=100)
    parser.add_argument("--background", default="dark")
    parser.add_argument("--region", action="append", default=[], type=parse_region)
    parser.add_argument("--save-normalised", help="also save the resized image, alpha intact, e.g. to composite onto later")
    args = parser.parse_args()

    width, height = (int(v) for v in args.size.split("x"))
    source = Image.open(args.image)
    source_aspect, target_aspect = source.width / source.height, width / height
    drift = abs(source_aspect - target_aspect) / target_aspect
    print(f"source {source.width}x{source.height} (aspect {source_aspect:.4f}), target {width}x{height} (aspect {target_aspect:.4f})")
    if drift > 0.01:
        print(f"WARNING: aspect differs by {drift:.1%} — likely player chrome or letterboxing in a screenshot. "
              "Crop to the video area first or every coordinate will be skewed.")

    normalised = source.convert("RGBA")
    if normalised.size != (width, height):
        normalised = normalised.resize((width, height), Image.LANCZOS)
    if args.save_normalised:
        normalised.save(args.save_normalised)

    flat = Image.new("RGBA", (width, height), parse_background(args.background) + (255,))
    flat.alpha_composite(normalised)
    flat = flat.convert("RGB")

    if args.region:
        crops = [gridded(flat, box, step, zoom, name) for name, box, step, zoom in args.region]
    else:
        crops = [gridded(flat, (0, 0, width, height), args.step, 1)]

    sheet = Image.new("RGB", (max(c.width for c in crops), sum(c.height for c in crops) + 10 * (len(crops) - 1)))
    y = 0
    for crop in crops:
        sheet.paste(crop, (0, y))
        y += crop.height + 10
    sheet.save(args.out)
    print(f"wrote {args.out}" + (f" and {args.save_normalised}" if args.save_normalised else ""))


if __name__ == "__main__":
    main()
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall