🎬 Faceless Viral Factory — Architecture

End-to-end faceless channel builder. YouTube API + AI video pipeline. Ollama (Mac Mini) for script generation. MiniMax API for TTS. FFmpeg for video assembly.

🔄 Production Pipeline

Step 1: Niche Selection

Agent researches CPM data, audience size, saturation. Outputs: 3 niche recommendations with data backing. Human approves one.

Step 2: Script Generation

For each video: 3 angle variations (emotional, surprising, educational). Ollama Mistral 7B on Mac Mini generates drafts. Human picks winner or requests revision.

Step 3: Voiceover

MiniMax TTS API or ElevenLabs converts script to audio. Background music from Epidemic Sound API (or free: Pixabay audio). Output: MP3.

Step 4: Visual Assembly

Pexels/Pixabay API fetches relevant stock footage based on script keywords. FFmpeg assembles: audio track + B-roll + captions + transitions. Output: MP4.

Step 5: Thumbnail

AI image generation (DALL-E or local SD) creates 3 thumbnail concepts. Human picks best.

Step 6: Upload + SEO

YouTube Data API uploads with title, description, tags (SEO-optimized by Ollama), scheduled publish time.

🛠️ Tech Stack

Component	Technology	Why
Script Generation	Ollama Mistral 7B (Mac Mini)	Local, fast, free per script
TTS	MiniMax TTS API	High quality, multiple voices, cheap per minute
Video Assembly	FFmpeg	Open source, command-line video editing
Stock Footage	Pexels API + Pixabay API	Copyright-free, API access, large library
Thumbnail Generation	DALL-E API or local Stable Diffusion	3 concepts per video, A/B testable
YouTube API	Google YouTube Data API v3	Upload, schedule, metadata
Hosting	Mac Mini + Google Cloud Storage	Rendered videos stored before upload

← Requirements | All Projects | Presentation →