AutoVideoProducer

A fully automated, multi-channel YouTube Shorts pipeline. Scrapes trending stories with GPT-4o, narrates them with OpenAI TTS, auto-syncs background gameplay footage via FFmpeg, renders with Remotion (React), and uploads to five YouTube channels on a rolling schedule — all driven from a single .env.

Independent project, originally built June 2025.

---

Architecture

          ┌────────────────────────── uploadManager ──────────────────────────┐
          │                                                                   │
          │   ScheduledExecutorService ─► ChannelScheduler ─► YouTubeUploader │
          │           ▲                         │                             │
          │           │ fills pending/ when     │ moves to archive/           │
          │           │ queue < 5 clips         ▼                             │
          │           │                    Channels/<name>/{pending,archive}  │
          └───────────┼───────────────────────────────────────────────────────┘
                      │ invokes
                      ▼
          ┌─────────────────────────── videoBuilder ──────────────────────────┐
          │   AIscraper (GPT-4o)  ─►  VidBuilder  ─► BackgroundGenerator      │
          │                           │                (FFmpeg)               │
          │                           ├─► WhisperTranscriber                  │
          │                           └─► Renderer ─► npx remotion render ───►│ .mp4
          └───────────────────────────────────────────────────────────────────┘
                                       ▲
                                       │ reads composition + captions
                                       │
                                    vidRenderer (Remotion / React)

Features

• Five YouTube channels driven in parallel with independent OAuth clients
• GPT-4o script generation with channel-specific prompts and history-aware de-duplication
• OpenAI shimmer TTS → MP3
• Whisper verbose-JSON transcription → word-timed captions
• FFmpeg-trimmed gameplay background at random start offsets
• Remotion (React) composition renders vertical 1080×1920 MP4
• 5-hour spacing per channel with automatic quota-rescheduling
• All secrets & paths read from a single project-root .env

Tech stack

Layer	Tool
AI script	OpenAI GPT-4o Responses API (AIscraper.java)
TTS + captions	OpenAI TTS + Whisper (whisper-1)
Video mixing	FFmpeg / ffprobe
Rendering	Remotion + React + Node.js
Upload	YouTube Data API v3 (OAuth2 desktop-app flow)
Scheduling	Java 17 ScheduledExecutorService
Build	Maven (multi-module)
Config	dotenv-java (.env in repo root)

Project structure

AutoVideoProducer/
├── pom.xml                         # Maven parent
├── .env / .env.example             # Secrets and path overrides
├── uploadManager/                  # Scheduler + YouTube upload
│   └── src/main/java/vid/manager/
│       ├── Main.java               # Entry point for the uploader loop
│       ├── ChannelScheduler.java   # Per-channel 5-hour scheduler
│       └── YouTubeUploader.java    # OAuth + resumable upload
├── videoBuilder/                   # Content generation
│   └── src/main/java/vid/builder/
│       ├── Main.java               # CLI: java vid.builder.Main <channel> <n>
│       ├── AIscraper.java          # GPT-4o + web_search_preview
│       ├── VidBuilder.java         # Orchestrates the whole pipeline
│       ├── WhisperTranscriber.java # speech.mp3 → timestamped segments
│       ├── BackgroundGenerator.java# FFmpeg clip + re-encode
│       ├── Renderer.java           # npx remotion render
│       ├── SearchResult.java       # Record (title, text)
│       ├── Env.java                # dotenv-java wrapper
│       └── AppPaths.java           # Centralised paths from .env
├── vidRenderer/                    # Remotion (React / TS) project
│   └── src/
│       ├── CaptionedShort.tsx      # The 1080×1920 composition
│       └── Root.tsx
└── Channels/<name>/                # Created at runtime per channel
    ├── pending/                    # Rendered MP4s awaiting upload
    ├── archive/                    # Uploaded MP4s
    └── upload_history.txt          # Timestamped log

Getting started

Prerequisites

• Java 17+ and Maven
• Node.js 18+ (for Remotion)
• FFmpeg + ffprobe on PATH
• An OpenAI API key
• YouTube Data API OAuth clients (one per channel) from the Google Cloud Console

Install

# 1. Clone
git clone https://github.com/JustinV200/AutoVideoProducer.git
cd AutoVideoProducer

# 2. Configure secrets
cp .env.example .env        # then fill in OPENAI_API_KEY and channel creds

# 3. Remotion deps
cd vidRenderer && npm install && cd ..

# 4. Build Java
mvn -q package

Run

Generate 3 videos for one channel:

java -cp videoBuilder/target/classes:$(mvn -q -f videoBuilder/pom.xml dependency:build-classpath -Dmdep.outputFile=/dev/stdout) \
     vid.builder.Main Channel_1 3

Start the long-running uploader loop (generates + schedules + uploads):

java -cp uploadManager/target/classes:$(mvn -q -f uploadManager/pom.xml dependency:build-classpath -Dmdep.outputFile=/dev/stdout) \
     vid.manager.Main

On Windows PowerShell, replace the classpath separator : with ;.

Configuration reference

All configuration lives in .env (see .env.example). Keys:

Key	Default	Purpose
OPENAI_API_KEY	(required)	Script, TTS, and Whisper calls
CHANNELS_ROOT	Channels	Per-channel output folders
VIDRENDERER_DIR	vidRenderer	Remotion project
GAMEPLAY_DIR	videoBuilder/src/main/resources/Gameplay_stores	Long gameplay source clips
NPX_PATH	npx	Override on Windows if needed
CHANNEL_<N>_CLIENT_ID / _SECRET / _EMAIL	(required)	OAuth credentials per channel

Real process environment variables override .env entries.

License

MIT