2noise/ChatTTS
ChatTTS is a generative text-to-speech model optimized for natural and expressive dialogue-based speech synthesis, supporting multi-speaker and fine-grained prosody control for conversational AI applications.
Awesome AI Agents › Audio & Voice Assistants
saharmor/voice-lab
Voice Lab is a comprehensive testing and evaluation framework designed specifically for voice agents powered by large language models (LLMs). It streamlines the process of evaluating and iterating on LLM-powered agents by providing tools to optimize costs, test new models, and fine-tune prompts for improved performance. Although optimized for voice agents, it is also valuable for evaluating any LLM-powered agents. The framework addresses common challenges in voice agent development such as manual review of call logs, lack of clear metrics for prompt refinement, risks associated with switching language models, and limited systematic testing of edge cases. Voice Lab enables users to define custom metrics in JSON format and employs an LLM-as-a-Judge approach to score these metrics, allowing for detailed performance tracking across different configurations. It supports confident model migration and cost optimization by facilitating comparisons between various models, including smaller, more efficient ones that balance cost and latency. The framework also supports systematic testing of multiple prompt variations and simulates diverse user interactions to verify performance. Installation involves cloning the repository, setting up a Python virtual environment, and configuring environment variables such as the OpenAI API key. Usage currently focuses on testing the text components of voice agents, with pre-defined tests available and advanced configuration supported through a dedicated configuration editor. Users can add new test scenarios by editing JSON configuration files, enabling customization of system prompts, initial messages, tool calls, success criteria, and agent personas. Future enhancements planned for Voice Lab include voice analysis features, support for additional language models, integration with persona datasets for complex scenarios, batch processing for cost reduction, and improved test reporting and execution capabilities. The project encourages contributions and provides attribution guidelines for users. Keywords: voice agents, large language models, LLM evaluation, prompt testing, model migration, cost optimization, custom metrics, JSON configuration, agent personas, performance testing, voice agent development, systematic testing, LLM-as-a-Judge, prompt refinement, test scenarios, voice analysis, batch processing, test reporting, open source framework.
https://github.com/saharmor/voice-lab
Also in Audio & Voice Assistants
neonbjb/tortoise-tts
Tortoise is a high-quality multi-voice text-to-speech system focused on realistic prosody and intonation, offering various usage modes and advanced performance optimizations.
livekit/agents
LiveKit Agents is an open-source framework for building real-time voice AI agents with integrated speech-to-text, large language models, text-to-speech, and telephony capabilities.
openinterpreter/01
01 is an open-source voice interface platform that enables natural language voice control across desktop, mobile, and ESP32 devices, offering extensive customization and hardware support.
ahmedeltaher/Android-MVVM-Architecture-Android-Voice-AI-SDK
A reusable Android library in Kotlin that adds a full voice conversation pipeline to an app, chaining microphone capture, voice activity detection, speech-to-text, an Anthropic Claude response and text-to-speech.
elevenlabs/ui
ElevenLabs UI is a component library built on shadcn/ui that provides customizable React components to help developers build multimodal agent and audio applications faster.
akdeb/ElatoAI
ElatoAI enables real-time AI speech interaction on Arduino ESP32 devices using OpenAI, Gemini, and Eleven Labs AI models with secure websocket communication and a web app for control.
bolna-ai/bolna
Bolna is an open-source end-to-end platform for building voice-first conversational AI agents by orchestrating telephony, speech recognition, language models, and speech synthesis technologies.