EVA is an open-source end-to-end framework for evaluating conversational voice agents across complete multi-turn spoken interactions. It measures both task accuracy (EVA-A) and conversation experience (EVA-X), using automated bot-to-bot audio simulations, validation, and metric scoring for cascade and audio-native systems.

This repository provides a lightweight text-to-speech API and CLI for Zhipu/GLM TTS, exposing speech synthesis endpoints and supporting built-in voices as well as cloned voices from audio.z.ai. It is packaged for easy deployment via Docker, Docker Compose, Home Assistant add-ons, and command-line usage.

Telnyx AI is a monorepo for building AI agents and AI-first developer workflows on top of Telnyx. It provides agent toolkits for Python and TypeScript, an agent-friendly CLI for provisioning telecom infrastructure, MCP support, unified plugins/extensions for coding assistants, a large catalog of agent skills, and operational guides for Telnyx APIs and services.

This repository is a PyTorch implementation of StyleSpeech / Meta-StyleSpeech, a multi-speaker adaptive text-to-speech system. It supports training, preprocessing, evaluation, and inference for speech synthesis with controllable pitch, energy, and duration, and includes HiFi-GAN-based vocoding for waveform generation.

LocalText2Voice is a desktop application for turning long-form text such as books, lessons, articles, and courses into narrated audio, MP3 audiobooks, and podcast-style mixes. It supports multiple local/offline and optional cloud TTS engines, includes voice management, text normalization, Whisper-based review, subtitle generation, and audio mixing with music, fades, ducking, and normalization.

NachoBot is a multi-platform AI virtual life/bot system designed to run across QQ, Discord, and Bilibili with a core-plus-adapter architecture. It emphasizes long-term memory, multimodal perception (ASR/VLM), TTS integration, and platform-specific interaction features such as live chat, voice channels, and livestream support.

Sokuji is a cross-platform real-time speech translation app for bilingual meetings, available as both an Electron desktop application and a Chrome/Edge browser extension. It supports two-way live translation with subtitles, virtual microphone/audio routing, and both cloud providers and fully offline on-device inference using WASM/WebGPU.

ReadAny is a local-first AI e-book reader for desktop and mobile. It supports AI chat grounded in the current book, semantic search, highlights and notes, text-to-speech, reading statistics, and WebDAV sync across devices, with broad format support and optional integration with multiple AI providers.

OpenLive is a local, open-source voice and vision layer for AI agents. It provides an on-device conversation pipeline with speech-to-text, turn detection, text-to-speech, barge-in, camera/screen vision, and support for both hosted model providers and coding agents via ACP.

This repository is an Android screen translation app for real-time on-screen text translation. It captures the screen, performs OCR locally or via cloud/local HTTP services, translates the text with LLMs or machine translation engines, and overlays the translated result on top of the original content or in a floating window, with optional text-to-speech reading.