youtube-shorts-pipeline

name: youtube-shorts-pipeline description: "Fully automated YouTube Shorts production pipeline. Takes a one-line news item or topic and outputs a finished, uploaded YouTube Short — complete with AI-generated b-roll visuals (Gemini Imagen), professional voiceover (ElevenLabs), SRT captions in English and Hindi, and direct YouTube upload via OAuth. Anti-hallucination research gate built in. Use when asked to create a YouTube Short, produce a news short, automate video content, run the content pipeline, draft/produce/upload a short-form video, or generate a faceless YouTube channel video."

YouTube Shorts Pipeline

Fully automated pipeline: news headline → research → script → AI visuals → voiceover → captions → upload.

First run triggers an automatic setup wizard — just run any command and you'll be guided through API key configuration and YouTube OAuth.

What it does

1. Draft — Researches topic via DuckDuckGo, writes Claude script, generates Gemini image prompts, YouTube description, Instagram caption, thumbnail prompt
2. Produce — Gemini Imagen generates 3 b-roll frames (Ken Burns animation), ElevenLabs synthesises voiceover, ffmpeg assembles final video with captions
3. Upload — Pushes to YouTube with title, description, tags, and SRT caption tracks (EN + HI)

Setup (one-time)

See references/setup.md for full instructions.

On first run, the pipeline automatically prompts for:

• ANTHROPIC_API_KEY — Claude script generation
• ELEVENLABS_API_KEY — voiceover (optional; macOS say fallback available)
• GEMINI_API_KEY — AI b-roll image generation

Config saved to: ~/.youtube-shorts-pipeline/config.json

Required binaries:

# macOS
brew install ffmpeg
pip install openai-whisper

# Linux
apt install ffmpeg
pip install openai-whisper

Required Python packages:

pip install anthropic google-api-python-client google-auth google-auth-oauthlib \
            pillow requests

Commands

Replace {skillDir} with the path where you extracted this skill.

Draft only

python3 {skillDir}/scripts/pipeline.py draft --news "your news headline here"

Produce from draft

python3 {skillDir}/scripts/pipeline.py produce --draft ~/.youtube-shorts-pipeline/drafts/<id>.json --lang en
python3 {skillDir}/scripts/pipeline.py produce --draft ~/.youtube-shorts-pipeline/drafts/<id>.json --lang hi

Upload

python3 {skillDir}/scripts/pipeline.py upload --draft ~/.youtube-shorts-pipeline/drafts/<id>.json --lang en
python3 {skillDir}/scripts/pipeline.py upload --draft ~/.youtube-shorts-pipeline/drafts/<id>.json --lang hi

Full auto (draft → produce → upload)

python3 {skillDir}/scripts/pipeline.py run --news "your news headline here"

Dry run (draft only, no video/upload)

python3 {skillDir}/scripts/pipeline.py run --news "your news headline here" --dry-run

Key Rules

• Anti-hallucination: Claude ONLY uses names/facts found in live DuckDuckGo research. Never fabricates player names, scores, or events.
• Hindi: Always native Hinglish writing, never translation. Fresh script for Hindi-speaking Indian audience.
• Whisper + Hindi: Whisper outputs Urdu script for Hindi audio — use Whisper timings + manually write Devanagari SRT.
• YouTube quota: uploadLimitExceeded = daily cap hit. Wait 24h.

File locations

• Config: ~/.youtube-shorts-pipeline/config.json
• Drafts: ~/.youtube-shorts-pipeline/drafts/<timestamp>.json
• Videos: ~/.youtube-shorts-pipeline/media/pipeline_<id>_<lang>.mp4
• SRTs: ~/.youtube-shorts-pipeline/media/pipeline_<id>_<lang>.srt

Voice IDs (ElevenLabs)

• English VO: JBFqnCBsd6RMkjVDRZzb (George — deep, authoritative)
• Override via env: VOICE_ID_EN=<id> or VOICE_ID_HI=<id>

Customising

• Channel topic: Pass any news item — esports, tech, finance, sports, politics
• Language: --lang en or --lang hi (or extend for any ElevenLabs-supported language)
• B-roll style: Edit image prompts in draft JSON before producing
• Script edits: Pass --script "edited script" to produce command to override Claude's draft

Troubleshooting

See references/troubleshooting.md for common errors and fixes.