TTS Studio — Compare and test various browser based TTS…

Pitch

Description

Comments

About this project

TTS Studio offers a unified web-based interface that allows users to easily test and compare multiple text-to-speech models. It features Kitten TTS, Piper TTS, and Kokoro TTS, all running in the browser, with seamless model switching and a variety of voice selections. Ideal for anyone looking to create high-quality speech output directly from their browser.

TTS Studio offers a powerful and user-friendly web-based interface for testing and comparing multiple text-to-speech (TTS) models directly in your browser. With seamless integration of Kitten TTS, Piper TTS, and Kokoro TTS, users can effortlessly select from a range of voice options tailored to various applications.

Key Features:

Triple Model Support:

Kitten TTS: A lightweight model at 24MB, delivering 8 expressive voice embeddings for efficient synthesis.
Piper TTS: A high-quality option with 75MB and an extensive selection of 904 diverse voices.
Kokoro TTS: A premium model of 82MB offering 21 expressive voices for natural speech synthesis.

Advanced Controls:

Model Switcher: Easily switch between different TTS engines to find the ideal voice.
Dynamic Interface: Tailors controls based on the model's capabilities.
Instant Voice Preview: Hear voice samples immediately by clicking the preview button.
Adjustable Speed: Control the speech rate from 0.5x to 2.0x to suit project requirements.
Sample Rate Management: Choose among various quality levels for audio output.
WebGPU Acceleration: Enable GPU support for faster processing where applicable.

Fully Browser-Based Operation:

TTS Studio functions entirely locally without needing a server.
Utilizes real-time generation powered by WebAssembly for rapid synthesis.
Loads only the selected model to optimize bandwidth and performance, featuring intelligent caching for speed.

Usage Example:

To generate speech with your chosen model, use the following commands:

    git clone https://github.com/clowerweb/tts-studio
    cd tts-studio
    npm install
    npm run dev

Then open in a browser and select your model to produce speech.

Technical Overview:

TTS Studio employs a robust technical stack, including Vue 3, Vite, and Tailwind CSS for the frontend, coupled with ONNX Runtime Web for machine learning capabilities. This setup enhances the performance and flexibility of text-to-speech generation.

Voice Comparison:

Feature	Kitten TTS	Kokoro TTS	Piper TTS
Model Size	~24MB	~82MB	~75MB
Voices Available	8 expressive	21 premium voices	904 diverse speakers
Voice Preview	✅ Instant	✅ Instant	✅ Instant
Speech Quality	High quality	Natural and expressive	Premium quality
Speed	~2-3x real-time	~1-2x real-time	~3-5x real-time
WebGPU	✅ Optional	✅ Optional	❌ WASM only
Sample Rate	✅ 8-48kHz configurable	24kHz fixed	22kHz fixed

Showcase Your Voice:

Instantly test voices with the one-click preview feature to ensure the selected voice meets your needs. With no complicated setup, users can easily evaluate various voice traits before making a selection.

Contribute:

The TTS Studio welcomes contributions to enhance functionality or add new features. Get involved by reporting issues, suggesting models, or submitting pull requests.

This project harnesses cutting-edge TTS technology to provide diverse and high-quality vocal outputs for developers, content creators, and anyone interested in advanced speech synthesis.