embed-subtitles

Embed Subtitles

Burn SRT subtitles into video files using FFmpeg.

Prerequisites

No SRT file? Use the transcribe skill first to generate subtitles from audio/video.

Need translation? After transcribing, read the SRT, translate each entry preserving timestamps, write new SRT file. No API needed.

Quick Start

cd ~/.claude/skills/embed-subtitles/scripts

# Burn SRT into video
npx ts-node embed-subtitles.ts -i video.mp4 -s subtitles.srt -o output.mp4

# Custom styling
npx ts-node embed-subtitles.ts -i video.mp4 -s subtitles.srt -o output.mp4 \
  --font-size 24 --position bottom --margin 40

# Add credit text overlay at end
npx ts-node embed-subtitles.ts -i video.mp4 -s subtitles.srt -o output.mp4 \
  --credit "Your Name" --credit-position top

Options

Option	Short	Default	Description
--input	-i	(required)	Input video file
--subtitles	-s	(required)	SRT subtitle file
--output	-o	(required)	Output video file
--font-size		20	Font size in pixels
--font-name		Arial	Font family name
--position		bottom	Subtitle position: top, center, bottom
--margin		25	Margin from edge in pixels
--color		white	Text color
--outline		2	Outline thickness
--shadow		1	Shadow depth
--credit			Credit text to show at end
--credit-position		top	Credit position: top, bottom
--credit-duration		2.5	Credit display duration in seconds
--extract-frame			Extract frame at specified second for verification

Position Values

• top - Near top of video (good for credits, avoids bottom burned-in text)
• center - Middle of video
• bottom - Near bottom of video (traditional subtitle position)

Dependencies

• FFmpeg (must be installed and in PATH)

RTL Auto-Detection & Fix (CRITICAL)

Before embedding ANY SRT file, Claude MUST check if the content is RTL.

Detection

Read the SRT file and check if the majority of text lines contain Hebrew/Arabic/Farsi characters:

• Hebrew range: \u0590-\u05FF
• Arabic range: \u0600-\u06FF
• Farsi additional: \u0750-\u077F

If RTL content is detected, apply the fix BEFORE passing to FFmpeg.

RTL Fix

Wrap each text line (not index numbers, not timestamps) with Unicode directional marks:

python3 -c "
import re

with open('SRT_FILE_PATH', 'r', encoding='utf-8') as f:
    content = f.read()

lines = content.split('\n')
result = []
for line in lines:
    stripped = line.strip()
    if stripped and not re.match(r'^\d+$', stripped) and not re.match(r'\d{2}:\d{2}:\d{2}', stripped):
        line = '\u202B' + line + '\u202C'
    result.append(line)

with open('SRT_FILE_PATH', 'w', encoding='utf-8') as f:
    f.write('\n'.join(result))
"

• U+202B (RLE - Right-to-Left Embedding): forces RTL paragraph direction
• U+202C (PDF - Pop Directional Formatting): closes the embedding

This fixes: English words at line start, periods/commas on wrong side, mixed bidi text.

When to apply

Claude should analyze the SRT content intelligently:

• If most text is Hebrew/Arabic/Farsi → apply RTL fix
• If mixed but predominantly RTL → apply RTL fix
• If LTR (English, Spanish, etc.) → skip
• No flag needed from the user - detect automatically

Notes

• For Arabic credits, use English text due to FFmpeg drawtext RTL limitations
• Use --extract-frame to verify subtitle positioning before final render