AlertTicker Card is a custom Home Assistant Lovelace frontend card for displaying and managing entity-based alerts and notifications. It provides extensive visual customization with dozens of animated themes, alert conditions, priorities, snoozing, history, TTS and push notifications, global overlay banners, camera support, music-player mode, weather widgets, multilingual support, and a full visual editor requiring no YAML.

StreamCore is a Go-based realtime media runtime for AI voice applications. It provides WebRTC audio over WHIP, NAT traversal, adaptive VAD and turn-taking, interruption handling, session management, realtime events, and streaming STT/LLM/TTS orchestration while allowing applications to supply their own agents, models, tools, and business logic. It supports browser, mobile, backend, telephony, and embedded-device clients.

KKClaw is a cross-platform Electron desktop companion that gives OpenClaw and Hermes AI agents a visible, interactive glass-orb avatar. It provides emotion and idle-expression animations, mouse interaction, configurable personalities, MiniMax voice cloning and TTS fallback, synchronized subtitles, model hot-swapping, multi-channel messaging, setup and diagnostic wizards, and automated Gateway monitoring and recovery.

floe-guard is a local, estimate-based spend meter and hard budget gate for AI agents. It prices LLM, tool, STT, TTS, and telephony usage from bundled cost maps, blocks calls or voice turns before they exceed a USD ceiling, supports concurrency-safe reservations, budgeting advisories, spend logs, and optional hosted ledger synchronization. It provides Python and TypeScript packages with adapters for OpenAI, Anthropic, Gemini, LangChain, LangGraph, CrewAI, LiteLLM, Vercel AI SDK, Pipecat, LiveKit, Vapi, and Retell.

BlahST is a lightweight, Linux-focused speech-to-text input toolkit that lets users dictate text into editable fields through configurable keyboard shortcuts. It uses whisper.cpp or whisperfile for offline and network transcription, supports multilingual transcription and English translation, continuous dictation, clipboard or automatic pasting, and experimental local-LLM speech assistants, proofreading, translation, and speech-to-speech chat with text-to-speech output.

ZhTTS is a Chinese text-to-speech demonstration designed to run in real time on CPU. It converts Chinese text into 24 kHz WAV audio using TensorFlow Lite FastSpeech2 or Tacotron2 acoustic models with an MB-MelGAN vocoder, and includes Chinese text normalization, pinyin/phoneme processing, punctuation pauses, a Python API, and an optional Flask web service.

Awesome Text-to-Speech is a curated directory of AI voice generation and speech synthesis resources. It catalogs commercial TTS platforms, open-source and local-first models, voice-cloning tools, research papers, demos, tutorials, community resources, and common use cases, with emphasis on multilingual, expressive, low-latency, and CPU-friendly solutions.

Voice Studio by MSR is a fully offline local web application for generating and editing speech with seven open-source text-to-speech engines, voice cloning/design, speech-to-text transcription, translation, and audio dubbing. It provides Podcast, Text-to-Voice, Transcribe, and Dub modes, with model switching, voice uploads, caching, subtitle export, and GPU/CPU/MPS support.

Kokoro Web is a browser-based text-to-speech application powered by the Kokoro 82M model and Transformers.js. It supports local speech generation using WebGPU or WASM, selectable voices, adjustable speech speed, streaming audio chunks, playback controls, text statistics, and WAV audio downloads.

SGLang-Omni is a multi-stage serving runtime for multimodal, speech, text-to-speech, music-generation, and audio-transcription models. It coordinates heterogeneous pipeline stages, specialized schedulers, inter-stage tensor transport, model integrations, and OpenAI-compatible APIs, with support for streaming, voice cloning, diarization, routing, and comprehensive performance and accuracy benchmarks.