Generate high-quality voice-over audio using OpenAI's Text-to-Speech API.
Available Voices
| Voice | Description | Best For |
|---|---|---|
| alloy | Neutral, balanced | General purpose |
| echo | Warm, conversational | Podcasts, casual |
| fable | Expressive, animated | Storytelling |
| onyx | Deep, authoritative | Professional, news |
| nova | Clear, friendly | Tutorials, explainers |
| shimmer | Soft, gentle | Meditation, calm content |
Usage
Command Line
node scripts/tts.js "Your text here" output/voiceover.mp3 --voice nova
With Different Voices
# Professional narrator
node scripts/tts.js "Welcome to our presentation" output/intro.mp3 --voice onyx
# Friendly tutorial
node scripts/tts.js "Let me show you how this works" output/tutorial.mp3 --voice nova
# Storytelling
node scripts/tts.js "Once upon a time..." output/story.mp3 --voice fable
From Script File
# Read script from file and generate audio
node scripts/tts.js "$(cat output/scripts/my-script.txt)" output/narration.mp3 --voice nova
Model Options
| Model | Quality | Speed | Cost |
|---|---|---|---|
| tts-1 | Good | Fast | Lower |
| tts-1-hd | Best | Slower | Higher |
Default: tts-1-hd (high quality)
Tips
Pacing
- Add periods for natural pauses
- Use commas for brief pauses
- Add "..." for longer pauses
Pronunciation
- Spell out acronyms: "API" → "A P I"
- Use phonetic spelling for unusual words
- Numbers are read naturally: "1000" → "one thousand"
Chunking Long Text
For scripts over 4096 characters:
- Split into logical chunks (paragraphs)
- Generate audio for each chunk
- Concatenate with ffmpeg
# Concatenate audio files
ffmpeg -f concat -safe 0 -i filelist.txt -c copy output/full-narration.mp3
# filelist.txt format:
# file 'chunk1.mp3'
# file 'chunk2.mp3'
Output
- Format: MP3
- Sample rate: 24000 Hz
- Channels: Mono
- Bitrate: 128 kbps (approximate)
Cost Estimate
~$0.015 per 1000 characters for tts-1-hd
| Script Length | Approx Cost |
|---|---|
| 500 chars | $0.0075 |
| 1000 chars | $0.015 |
| 5000 chars | $0.075 |