🎥 Course Pipeline

Automated Google Slides → Narrated Videos → Final Course Video

course_pipeline is a deterministic, reproducible, dependency-aware video generation pipeline that converts a Google Slides deck into a fully narrated course video.

It supports:

• Human voice recording
• macOS native TTS (say)
• ElevenLabs high-quality TTS
• Incremental rebuilds
• Slide-level redo
• Fully ordered, dependency-driven execution using doit

This is designed for course creators, educators, and AI-first content pipelines.

---

✨ Features

• 🔗 Google Slides → Video (single source of truth)

• 🧠 Dependency-aware pipeline (no reruns unless needed)

• 🔢 Strict numeric ordering (1 → N slides, no randomness)

• 🎙️ Multiple audio modes:

  • Interactive human recording
  • macOS say
  • ElevenLabs API (studio-quality voices)

• 🎞️ Per-slide video generation

• 🎬 Final concatenation into final.mp4

• ♻️ Incremental & resumable (crash-safe)

• 🧹 Redo individual slides easily

---

📦 Pipeline Overview

Google Slides URL
        │
        ▼
extract_notes     → notes/1.txt ... notes/N.txt
screenshot        → screenshots/1.png ... N.png
audio / tts_*     → audio/1.wav ... N.wav
video             → videos/1.mp4 ... N.mp4
final             → final.mp4

Each step is a doit task with explicit dependencies.

---

🛠️ Installation

Prerequisites

• Python 3.11+
• macOS (for say; optional if using ElevenLabs)
• ffmpeg
• Google Cloud credentials (Slides + Drive API)

Install ffmpeg:

brew install ffmpeg

Clone & setup

git clone https://github.com/<your-org>/course_pipeline.git
cd course_pipeline
python -m venv .venv
source .venv/bin/activate
pip install uv
uv sync

---

🔐 Credentials Setup

Google APIs

Enable:

• Google Slides API
• Google Drive API

Create OAuth credentials and save as:

credentials.json

First run will open a browser for auth and cache tokens locally.

---

ElevenLabs (optional but recommended)

export ELEVENLABS_API_KEY="your_api_key_here"

---

⚙️ Configuration

Edit settings in:

scripts/settings.py

Typical options:

• Video resolution (e.g. 1920x1080)
• FPS
• Output directories

---

🚀 Usage

Run the full pipeline

uv run doit

This will:

1. Extract notes
2. Take slide screenshots
3. Generate audio
4. Create per-slide videos
5. Concatenate into final.mp4

---

🎙️ Audio Options

1️⃣ Interactive Human Recording

uv run doit audio

• Displays slide text

• Records mic input

• Options:

  • Enter → save & continue
  • r → re-record
  • s → skip

---

2️⃣ macOS Text-to-Speech

uv run doit tts_slides

Uses native say command.

---

3️⃣ ElevenLabs TTS (recommended)

uv run doit elevenlabs_tts

• Studio-quality voices
• Handles empty slides safely
• Converts MP3 → WAV automatically

Force regeneration:

uv run doit elevenlabs_tts --force

---

🎞️ Video Generation

Generate per-slide videos:

uv run doit video

Each slide:

1.png + 1.wav → 1.mp4

Optimized for:

• QuickTime
• VLC
• YouTube compatibility

---

🎬 Final Output

uv run doit final

Result:

build/<deck_id>/final.mp4

Force re-encode (if needed):

uv run doit final --reencode

---

♻️ Redo a Single Slide

Redo slide 12 completely:

uv run doit redo --slide 12
uv run doit

Redo only audio:

uv run doit redo --slide 12 --what audio

---

📋 Available Tasks

doit list

Example:

setup
extract_notes
screenshot
audio
tts_slides
elevenlabs_tts
video
final
redo

---

🧠 Why doit?

This pipeline does not rely on ad-hoc scripts.

doit gives:

• True dependency graphs
• Incremental builds
• Crash-safe resumption
• Deterministic ordering

This makes it suitable for large decks (100+ slides) and CI-like automation.

---

🧩 Design Principles

• Single source of truth: Google Slides
• Artifacts are explicit: text, image, audio, video
• No hidden state
• No silent reprocessing
• Human-friendly override at every stage

---

📄 License

MIT License Free to use, modify, and distribute.

---

🤝 Contributing

PRs welcome for:

• Windows/Linux TTS backends
• Speaker diarization
• Subtitle (SRT/VTT) generation
• Video transitions
• Parallel execution

---

🙌 Acknowledgements

• Google Slides API
• FFmpeg
• ElevenLabs
• doit task runner