How to add boxed captions with Claude

Add captions to input.mp4 using the word timings in words.json. Show four words at a time as black text on a white rounded box, centered 72% of the way down, from the group's first word to its last word. Use Python with Pillow to draw it and ffmpeg to put it on the video (don't rely on ffmpeg's drawtext filter), keep the audio, and save it as output.mp4.

#!/usr/bin/env python3
"""Boxed captions: each phrase in black text on a white rounded box, on screen while it is spoken.

Usage: python3 boxed-captions.py input.mp4 words.json output.mp4
words.json holds word timings, for example from Whisper:
  [{"word": "This", "start": 0.30, "end": 0.69}, {"word": "serum", "start": 0.77, "end": 1.13}, ...]
Needs ffmpeg and Pillow (pip install pillow).
"""
import json
import math
import subprocess
import sys
from pathlib import Path

from PIL import Image, ImageDraw, ImageFont

SIZE = 0.065     # text height as a share of the frame width (shrinks to fit long phrases)
Y = 0.72         # where the box's center sits (0 = top, 1 = bottom)
CHUNK = 4        # words per phrase
PAD = 0.03       # space inside the box, as a share of the frame width
FONTS = ["/System/Library/Fonts/Supplemental/Arial Bold.ttf",     # macOS
         "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf",  # Linux
         "C:/Windows/Fonts/arialbd.ttf"]                          # Windows
FONT_FILE = next((f for f in FONTS if Path(f).exists()), None) or sys.exit("No bold font found: add one to FONTS")


fonts = {}


def fitted_font(text):
    size = round(w * SIZE)
    while size > 12:
        font = fonts.get(size) or fonts.setdefault(size, ImageFont.truetype(FONT_FILE, size))
        if font.getlength(text) <= w * (0.9 - 2 * PAD):
            return font
        size -= 2
    return font


def overlay(t):
    img = Image.new("RGBA", (w, h))
    for group in (words[i:i + CHUNK] for i in range(0, len(words), CHUNK)):
        if group[0]["start"] <= t < group[-1]["end"]:
            text = " ".join(word["word"].strip() for word in group)
            font = fitted_font(text)
            draw = ImageDraw.Draw(img)
            x0, y0, x1, y1 = draw.textbbox((w / 2, h * Y), text, font=font, anchor="mm")
            pad = w * PAD
            draw.rounded_rectangle([x0 - pad, y0 - pad, x1 + pad, y1 + pad], radius=pad, fill="white")
            draw.text((w / 2, h * Y), text, font=font, anchor="mm", fill="black")
    return img


def probe(path):
    info = json.loads(subprocess.check_output(["ffprobe", "-v", "error", "-select_streams", "v:0", "-show_entries",
                                               "stream=width,height,r_frame_rate:format=duration", "-of", "json", path]))
    s = info["streams"][0]
    num, den = s["r_frame_rate"].split("/")
    return s["width"], s["height"], s["r_frame_rate"], float(num) / float(den), float(info["format"]["duration"])


def render(src, out):
    """Draw overlay(t) on every frame of src and save to out, keeping the audio."""
    ff = subprocess.Popen(["ffmpeg", "-y", "-v", "error", "-i", src,
                           "-f", "rawvideo", "-pix_fmt", "rgba", "-s", f"{w}x{h}", "-r", rate, "-i", "-",
                           "-filter_complex", "[0:v][1:v]overlay=shortest=1,format=yuv420p[v]",
                           "-map", "[v]", "-map", "0:a?", "-c:v", "libx264", "-crf", "18", "-c:a", "copy", out],
                          stdin=subprocess.PIPE)
    for i in range(math.ceil(duration * fps)):
        ff.stdin.write(overlay(i / fps).tobytes())
    ff.stdin.close()
    if ff.wait():
        sys.exit("ffmpeg failed")


src, words_file, out = sys.argv[1:4]
words = json.loads(Path(words_file).read_text())
w, h, rate, fps, duration = probe(src)
render(src, out)