Loading repository data…
thisisanoopshekhar89-jpg / repository
Autonomous AI agent that runs a daily YouTube channel — news scanning to video production to SEO upload. Zero human editing. Claude Opus + ElevenLabs + FFmpeg + YouTube API.
A transparent discovery signal based on current public GitHub metadata.
This score does not audit code, security, maintainers, documentation quality, or suitability. Verify the repository and its current documentation before adoption.
python main.py— That's it. The pipeline does everything else.
A fully automated daily YouTube Shorts show — The Dumb Money Club — powered entirely by AI.
Two animated characters, Bull and Bear, present the day's biggest financial, political and viral news as a live TV show. Five stories. Five chapters. Five shorts. One main episode. All published daily without any human editing.
Bull is always wrong. Bear is always right. The comedy is watching Bull figure it out in real time.
python main.py
│
├── STEP 1 📰 News Scanner
│ NewsAPI + CoinGecko + Reddit
│ 5 unique stories — Global | Person | Market | Viral | Crypto
│
├── STEP 2 ✍️ Script Writer (Claude Opus)
│ One flowing TV show conversation
│ 5 chapters with CTAs and transitions
│ Raina energy — raw, unfiltered, true
│
├── STEP 2b 🖼️ Thumbnail Generator
│ Character PNGs composited on dark background
│ Episode-specific text from script SEO field
│
├── STEP 3 🎙️ Voice Engine (ElevenLabs)
│ Bull, Bear and Announcer voiced
│ All beats — cold_open, argument, CTA, transition
│
├── STEP 4 🎬 Video Assembler
│ Characters lip-synced to every beat
│ Expression PNG swapped per dialogue beat
│
├── STEP 5 🎵 Intro + Outro
│ 12s animation with jingle and announcer
│ Branded outro slide
│
├── STEP 6 🔗 Final Assembly (FFmpeg)
│ Intro + Main + Outro concatenated
│
├── STEP 7 ✂️ Shorts Cutter
│ 5 shorts cut at exact story boundaries
│ Each short = cold_open → CTA (self-contained arc)
│
├── STEP 8 🔍 QC Bot
│ Automated quality check before upload
│ Blocks upload if standards not met
│
└── STEP 9 📺 YouTube Uploader
Main episode + 5 shorts
SEO titles, descriptions, tags, thumbnail
Curiosity gap title format proven by analytics
Selected from shared topics, language and repository description—not editorial ratings.
gijs-hulsebos /
AI Automation Engineer specializing in autonomous agentic workflows. Bridging the gap between LLM reasoning and business logic using n8n, Python, and OpenRouter. Currently building enterprise-grade tools like the Autonomous Calendar Assistant. Visit gijshulsebos.com for more
49/100 healthEach episode is structured as one continuous TV show — not five separate segments.
[INTRO — 12s animation + jingle + announcer]
↓
Chapter 1 — Global News
cold_open → bull → bear → argument → CTA → transition
↓
Chapter 2 — Name Said WHAT
cold_open → bull → bear → argument → CTA → transition
↓
Chapter 3 — Market Update (BTC / BNB / Gold)
cold_open → bull → bear → argument → CTA → transition
↓
Chapter 4 — Viral / AI / Startup
cold_open → bull → bear → argument → CTA → transition
↓
Chapter 5 — The Closer (philosophical truth)
cold_open → bull → bear → argument → Bear CTA only
↓
[OUTRO — jingle slide]
Shorts are extracted from each chapter — cold_open → CTA. The transition belongs to the main show only. Each short has a proper beginning, middle and end.
| Component | Technology | Purpose |
|---|---|---|
| Script | Claude Opus (Anthropic) | Full episode as one TV show |
| Voices | ElevenLabs | Bull, Bear, Announcer |
| Video | FFmpeg + Python | Assembly, cuts, audio mixing |
| News | NewsAPI + CoinGecko + Reddit | 5 unique daily stories |
| Upload | YouTube Data API v3 | Main + 5 shorts with SEO |
| Thumbnails | Pillow | Character PNG compositing |
| QC | Custom QC Bot | Pre-upload quality gate |
| Pipeline | Python 3.14 | Orchestration |
Based on channel analytics, unfinished titles outperform complete ones:
| Chapter | Title Format |
|---|---|
| Global | [fact]... and nobody saw this coming |
| Person | [fact]... Bear has the receipt |
| Market | [price/impact]... your wallet already knows |
| Viral | [hook]... Bear is not impressed |
| Closer | Bear said something tonight we cannot stop thinking about |
463 — The government found $1.5 trillion. It's not for your hospital.
438 — America owes the planet $10 trillion. Guess who's actually paying it.
350 — Someone on cable news suggested taxing the route that carries...
350 — Ford just spent millions building a car that costs more than your house.
302 — Scientists built a battery that charges a MILLION times faster...
Scripts are written in the spirit of sharp, unfiltered financial commentary:
dumbmoney/
├── main.py # Master pipeline — run this
├── news_scanner.py # 5-category news fetcher
├── script_writer.py # Claude Opus script writer
├── voice_engine.py # ElevenLabs voice generator
├── video_assembler.py # Character animation + lip sync
├── youtube_uploader.py # YouTube API uploader
├── thumbnail_generator.py # Auto thumbnail generator
├── qc_bot.py # Pre-upload quality control
├── twitter_poster.py # Twitter/X auto posting
└── sound_engine.py # Audio processing helpers
assets/
├── characters/
│ ├── bull/bull_v2/ # 20 Bull expression PNGs
│ └── bear/bear_v2/ # 20 Bear expression PNGs
├── stage/
│ └── introtoclub.mp4 # Intro animation
└── sounds/
├── jingle.mp3 # Intro jingle
└── jingleoutro.mp3 # Outro jingle
output/
├── episodes/YYYYMMDD/ # Generated episode files
├── news/ # Daily news scans
├── thumbnails/ # Generated thumbnails
└── episode_memory.json # Rolling 30-episode memory
pip install anthropic elevenlabs requests python-dotenv pillow
pip install google-api-python-client google-auth-oauthlib
FFmpeg must be installed and in PATH.
Create .env in the project root:
ANTHROPIC_API_KEY=your_key_here
ELEVENLABS_API_KEY=your_key_here
NEWSAPI_KEY=your_key_here
youtube_credentials.json to project rootpython -c "from youtube_uploader import get_youtube_service; get_youtube_service()"
Place character expression PNGs in:
assets/characters/bull/bull_v2/*.png
assets/characters/bear/bear_v2/*.png
Required expressions: neutral, excited_arms_up, adbhuta, raudra, hasya, smug, facepalm, thinking, pointing, arms_crossed, shringara, vira, karuna, shanta, bhayanaka, mic_drop, exaggerated_laugh, mouth_open, mouth_half, mouth_closed
# Full daily episode — news to upload
python main.py
# Test news scanner only
python news_scanner.py
# Test script writer only
python script_writer.py
# QC check on today's episode
python qc_bot.py
YouTube Shorts algorithm works in waves. Channels at this stage are tested with small audiences. One short breaking out resets the trajectory — all older content gets pushed simultaneously.
The content formula is already validated by the data. Curiosity gap titles, personal financial impact and the Bull vs Bear conflict drive watch time and shares.
Target: One short crossing 5,000 views by Day 60. Based on current trajectory and improving content quality, achievable without paid promotion.
Why one script for the full episode? Claude Opus writes all 5 chapters in one API call — it knows all the stories simultaneously and writes transitions that grammatically connect to the next cold_open. This makes the full episode feel like one conversation, not 5 stitched segments.
Why story-boundary shorts?
Shorts cut at equal time splits miss story beginnings and endings. We map actual beat video durations to find exact cold_open_start and cta_end timestamps, giving each short a proper arc.
Why QC bot before upload? Automated content can fail silently — corrupt audio, duplicate stories, missing expressions. The QC bot catches these before they go public.
Why curiosity gap titles? Channel analytics prove it. Every top-performing title cuts before the reveal. The viewer must click to find out what happened.
Built by a solo founder in 16 days. Zero employees. Zero budget. One command.
Channel: @DumbMoneyClub2026
Smart Money. Stupid World.