batch-processor

name	batch-processor
description	Download, transcribe, and clip multiple YouTube stream VODs by date range. Orchestrates the full pipeline from video download through clipping. Use when processing multiple videos at once.

Batch Video Processor Skill

Downloads stream VODs from a YouTube channel by date range, transcribes them, and runs the clipper skill on each one automatically.

Default behavior: Fetches from the /streams tab (VODs only), not regular uploads.

Quick Start

User: "download and clip all of my vids from dec 16 to dec 21"

You will automatically:

1. List videos from the channel in that date range
2. For each video: download → transcribe → clip
3. Organize output in per-video folders
4. Report summary when complete

Configuration

Default Channel: https://www.youtube.com/@techfren

Shell Aliases Used (from user's ~/.zshrc):

• yt = yt-dlp -f "bv*+ba/b" - downloads best video+audio
• tr = transcribe-anything --prompt_file ~/vocab.txt --language en - transcribes with custom vocabulary

Execution Checklist

When user requests batch processing, follow this exact sequence:

Phase 1: Parse Request & List Videos

• Parse date range from user's request (e.g., "dec 16 to dec 21" → 2024-12-16 to 2024-12-21)
• If year not specified, assume current year (or most recent occurrence)
• Run: python .claude/skills/batch-processor/scripts/list_channel_videos.py --start <start> --end <end> --json
• Parse JSON output to get list of videos
• Report to user: "Found N videos from [date range]"

Phase 2: Process Each Video

For each video in the list (process in chronological order):

A. Create Working Directory

# Format: YYYY-MM-DD_Sanitized_Title
mkdir -p "20241216_Video_Title_Here"
cd "20241216_Video_Title_Here"

Sanitize title by:

• Replacing /\:*?"<>| with _
• Removing non-alphanumeric except spaces, hyphens, underscores
• Truncating to 50 characters
• Replacing spaces with _

B. Download Video

yt "<video_url>"

• The yt alias downloads best quality video+audio
• Output filename will be auto-generated by yt-dlp (usually <title> [<id>].mp4 or similar)
• Detect the downloaded video file: ls *.mp4 *.mkv *.webm 2>/dev/null | head -1

C. Transcribe Video

tr "<video_file>"

• The tr alias runs transcribe-anything with the user's vocab file
• Output: out.json in current directory
• Wait for completion (can take several minutes for long videos)

D. Run Clipper Skill

Follow the clipper skill workflow:

1. Parse transcription (if parsed.json doesn't exist):

   python .claude/skills/clipper/scripts/parse_transcription.py out.json > parsed.json
   

2. Analyze for clips: Read parsed.json and identify segments following clipper skill instructions. Write segments.json.

3. Extract clips:

   python .claude/skills/clipper/scripts/extract_clips.py segments.json out.json <video_file> clips/
   

4. Cleanup clips:

   python .claude/skills/clipper/scripts/cleanup_clips.py segments.json out.json <video_file> clips/
   

E. Return to Parent Directory

cd ..

Phase 3: Summary Report

After all videos are processed, provide a summary:

Batch Processing Complete!
==========================

Date Range: Dec 16 - Dec 21, 2024
Videos Processed: 5

Results:
  1. 2024-12-16_Morning_Stream/
     - Duration: 2h 15m
     - Clips found: 23
     - Clips extracted: 18

  2. 2024-12-17_Tech_Tutorial/
     - Duration: 45m
     - Clips found: 8
     - Clips extracted: 6

  ... (etc)

Total:
  - Videos: 5
  - Total clips found: 67
  - Total clips extracted: 52
  - Skipped (too short): 12
  - Skipped (too long): 3

Output Structure

./
├── 2024-12-16_Morning_Stream/
│   ├── Morning Stream [abc123].mp4    # Original video
│   ├── out.json                        # Raw transcription
│   ├── parsed.json                     # Parsed sentences
│   ├── segments.json                   # Identified clips
│   └── clips/
│       ├── 001_OAuth_Setup.mp4
│       ├── 002_Redis_Config.mp4
│       └── cleaned/
│           ├── 001_OAuth_Setup.mp4
│           └── 002_Redis_Config.mp4
├── 2024-12-17_Evening_Stream/
│   ├── ...
└── 2024-12-18_Tutorial/
    └── ...

Date Parsing Examples

The script handles various date formats:

Input	Interpreted As
dec 16 to dec 21	2024-12-16 to 2024-12-21
December 16 to 21	2024-12-16 to 2024-12-21
12/16 to 12/21	2024-12-16 to 2024-12-21
last 7 days	Use --days 7 flag
past week	Use --days 7 flag
yesterday to today	Yesterday to today
2024-12-16 to 2024-12-21	Explicit dates

Error Handling

• No videos in range: Report "No videos found" and exit gracefully
• Download fails: Log error, skip to next video, include in summary
• Transcription fails: Log error, skip to next video, include in summary
• Clipping fails: Still have video + transcription, note partial completion
• Disk space: Check available space before starting (videos can be large)

Customization

Different Channel

User: "download and clip videos from @otherchannel from dec 16 to dec 21"

Use: python .../list_channel_videos.py "https://youtube.com/@otherchannel" --start ...

Skip Transcription (Already Have It)

User: "just clip the videos, I already have transcriptions"

Skip the tr step, look for existing out.json files.

Skip Download (Local Videos)

User: "transcribe and clip all mp4s in this folder"

Skip listing/download, iterate over *.mp4 files instead.

Scripts Reference

list_channel_videos.py

Lists YouTube channel stream VODs (or videos) filtered by date.

# Basic usage - stream VODs from default channel (default)
python .claude/skills/batch-processor/scripts/list_channel_videos.py --start "dec 16" --end "dec 21"

# JSON output for parsing
python .claude/skills/batch-processor/scripts/list_channel_videos.py --start "dec 16" --end "dec 21" --json

# Regular uploads instead of streams
python .claude/skills/batch-processor/scripts/list_channel_videos.py --videos -s "dec 16" -e "dec 21"

# Different channel
python .claude/skills/batch-processor/scripts/list_channel_videos.py "https://youtube.com/@channel" -s "dec 16" -e "dec 21"

# Last N days
python .claude/skills/batch-processor/scripts/list_channel_videos.py --days 7 --json

Arguments:

• channel: YouTube channel URL (default: https://www.youtube.com/@techfren)
• --start, -s: Start date
• --end, -e: End date
• --days, -d: Alternative: last N days
• --max, -m: Max videos to scan (default: 100)
• --json, -j: Output as JSON for parsing
• --streams: Fetch stream VODs only (default: True)
• --videos: Fetch regular uploads instead of streams

JSON Output Format:

{
  "channel": "https://www.youtube.com/@techfren",
  "date_range": {
    "start": "2024-12-16",
    "end": "2024-12-21"
  },
  "videos": [
    {
      "id": "abc123",
      "title": "Morning Stream",
      "upload_date": "20241216",
      "url": "https://www.youtube.com/watch?v=abc123",
      "duration_seconds": 8100
    }
  ]
}

Requirements

• yt-dlp: For listing and downloading videos (brew install yt-dlp)
• transcribe-anything: For transcription (already configured as tr alias)
• ffmpeg: For clip extraction (brew install ffmpeg)
• Python 3.8+: For all scripts

Integration with Other Skills

This skill orchestrates:

1. batch-processor (this skill): Download + transcribe
2. clipper: Analyze + extract + cleanup clips
3. media-fetcher (optional): Fetch logos/GIFs for clips after clipping

To also fetch media after clipping, user can say:

"download and clip all my vids from dec 16 to dec 21, then fetch media for the clips"