Gen-TTS: Gemini Native Audio Generation CLI

gen-tts is a powerful command-line interface for Google's Gemini Native Text-to-Speech (TTS) capabilities. It allows you to generate high-quality, expressive speech from text, including single-speaker narration, multi-speaker conversations, and AI-generated podcasts and summaries.

Powered by the Gemini 2.5 and Gemini 2.0 models.

Features

• High-Quality Voices: Access Gemini's full range of expressive voices (e.g., Charon, Kore, Fenrir, Puck).
• Multi-Speaker Support: Generate conversations between different speakers with distinct voices.
• Podcast Mode: Automatically turn any text or file into a lively "Deep Dive" podcast conversation between two hosts.
• Summary Mode: Summarize long text into a concise, information-packed audio briefing read by a warm, professional voice.
• Transcript Generation: Ask Gemini to write a script for you based on a topic, then immediately synthesize it.
• MP3 Support: Automatically converts output to MP3 (requires ffmpeg) or WAV.
• Cross-Platform Playback: Automatically plays the generated audio on Linux, macOS, and Windows.

Installation

Prerequisites

• Python 3.9+
• ffmpeg (Required for MP3 support and playback on Linux)
  • Linux: sudo apt install ffmpeg
  • macOS: brew install ffmpeg

Install via uv (Recommended)

uv tool install git+https://github.com/charles-forsyth/generate-tts.git

Install via pip

pip install git+https://github.com/charles-forsyth/generate-tts.git

Configuration

The tool requires a Google Cloud API Key. On first run, it will create a config file at ~/.config/gen-tts/.env where you can paste your key.

# ~/.config/gen-tts/.env
GOOGLE_API_KEY="your_actual_api_key_here"

Usage

1. Basic Single Speaker

Generate audio from text. Default voice is Charon (Deep, Warm Male).

gen-tts "System systems operational." --temp

--temp plays the audio immediately without saving a file.

2. Podcast Mode ("Deep Dive")

Turn an article, report, or text into an engaging podcast conversation between two hosts (Fenrir and Leda).

# From a file
gen-tts --input-file article.txt --podcast --output-file deep_dive.mp3

# Piping text
echo "Breaking news..." | gen-tts --podcast

3. Summary Mode

Summarize text into a concise, professional audio briefing. Default voice is Charon.

cat report.txt | gen-tts --summary --output-file briefing.mp3

4. Topic-Based Generation

Ask Gemini to write a script for you and then speak it.

gen-tts --generate-transcript "A funny debate about coffee vs tea" \
        --multi-speaker --speaker-voices Alice=Kore Bob=Puck \
        --output-file debate.mp3

5. Custom Multi-Speaker

Provide your own script formatted as Speaker: Text.

script.txt:

Joe: Hey Jane, did you see the update?
Jane: Yes, it looks amazing!

Command:

gen-tts --input-file script.txt --multi-speaker \
        --speaker-voices Joe=Charon Jane=Puck \
        --audio-format MP3

Options Reference

Flag	Description
--podcast	Generate a multi-speaker podcast script from input.
--summary	Generate a concise summary script from input.
--generate-transcript "TOPIC"	Generate a script based on a topic.
--multi-speaker	Enable multi-speaker mode (requires --speaker-voices).
--speaker-voices	Map speakers to voices (e.g., Host=Fenrir Guest=Leda).
--voice-name	Voice for single-speaker mode (Default: Charon).
--audio-format	WAV or MP3. Defaults to MP3 for podcasts/summaries.
--model	TTS Model (Default: gemini-2.5-flash-preview-tts).
--transcript-model	Model for script generation (Default: gemini-2.5-pro).
--list-voices	List all available Gemini voices.
--no-play	Disable automatic playback.

License

MIT License