Pitch Narration Generator

Parses a structured video script, extracts all Narrator: blocks, and synthesises them into a single MP3 using Azure OpenAI TTS.

How it works

1. Split the script on Time: ... section headings
2. Extract each Narrator: block
3. Synthesise each block via Azure OpenAI TTS → output/section_NN.mp3
4. Concatenate with a short pause between sections → output/narration.mp3

Setup

1. Install dependencies

pip install openai python-dotenv pydub

pydub requires ffmpeg for MP3 export:

brew install ffmpeg   # macOS

2. Configure credentials

cp .env.example .env

Fill in your Azure OpenAI TTS resource key and endpoint.

Usage

python video_gen.py script.txt

Your script file should use this format:

Time: 0:00-0:30 | Section Title
(Visual: optional stage direction — ignored)
Narrator: The text that will be spoken aloud.

Time: 0:30-1:00 | Next Section
Narrator: More narration here.

Output

output/
├── section_00.mp3   ← per-section audio clips
├── section_01.mp3
│   …
└── narration.mp3    ← final stitched narration

Configuration

Setting	Where
TTS voice	voice= in synthesise() — options: alloy, echo, fable, nova, onyx, shimmer
TTS speed	speed= in synthesise()
Pause between sections	PAUSE_MS constant