Podvoice is a local-first, open-source command line interface (CLI) tool that transforms simple Markdown scripts into captivating multi-speaker audio using Coqui XTTS v2. It is tailored for developers and content creators who seek a straightforward method to produce podcast-style audio without relying on cloud services or expensive APIs.

Key Features

• Markdown-Based Scripts: Easily create content as a .md file with designated speaker blocks.
• Multiple Logical Speakers: Each speaker is consistently mapped to a voice in the XTTS model, enabling smooth transitions.
• Single Output File: Podvoice compiles the entire script into one cohesive audio file, simplifying distribution.
• WAV or MP3 Export: The tool supports WAV format by default while allowing MP3 output if specified.
• Local-Only Inference: Operations are performed locally using a pre-trained Coqui XTTS v2 model that is downloaded once and cached for future use.
• CPU-Friendly Performance: Designed to run efficiently on CPU, with the option to utilize a GPU if available.
• Accessible Code: A modular Python 3.10+ codebase with comprehensive comments, making it user-friendly for beginners.

Example Usage

Podvoice expects scripts formatted in Markdown to define speakers and their corresponding dialogue:

[SpeakerA | calm]
Hello and welcome to the show.

[SpeakerB | excited]
Aaj hum AI ke baare mein baat karenge.

Quick Start

The main command to generate the audio output from a Markdown script is:

podvoice render SCRIPT.md --out OUTPUT

For instance:

podvoice render examples/demo.md --out output.wav

With explicit options to specify language and device:

podvoice render \
  examples/demo.md \
  --out podcast.mp3 \
  --language en \
  --device cpu

Responsible Use

Users are encouraged to use the Podvoice tool responsibly, refraining from any use that may infringe on the rights or dignity of others.

In summary, Podvoice provides a robust, efficient, and user-friendly solution for generating audio content from Markdown scripts, supporting a wide range of applications in the field of podcasting and conversational content.