Local AI for Speech, Voice and Music

MusicGen is Meta AI's music generation model within AudioCraft, a PyTorch library for developers and audio researchers who want to generate music in their own computing environment. It creates music from text descriptions and can use a melody to guide the result. AudioCraft includes both inference and training code, so it's suited to people building audio tools or studying music generation.

Unmute adds spoken conversation to text LLMs using Kyutai's speech recognition and speech synthesis models. It's for developers who want a self-hosted voice interface while keeping their choice of language model. The project uses the MIT license, and a hosted browser demo is available at Unmute.sh.

Moonshine is an on-device AI toolkit for developers building voice agents and applications that listen and speak. It combines speech to text, intent recognition and text to speech in one library. Voice processing stays on the device, and you don't need an account or API keys.

SenseVoice is a local speech recognition model that adds language, emotion and sound-event tags to transcriptions. It's for developers building voice applications or analyzing recordings on their own hardware, particularly those working with Mandarin and Cantonese. The project is open source under the MIT license.

MeloTTS is a Python text-to-speech library for developers who want to generate speech locally, including on machines without a dedicated GPU. It supports real-time inference on a CPU. Its language and accent choices make it relevant for applications that need spoken output across different audiences.

MacWhisper is a native macOS transcription app for people working with interviews, lectures, meetings and other recorded audio. It runs speech recognition on your own Mac, so local transcription keeps audio on your device. It also offers cloud transcription through services such as OpenAI, ElevenLabs and Deepgram, which send audio off your machine.

VideoLingo is a self-hosted video translation app for creators and educators who need bilingual subtitles or dubbed versions of their videos. It brings transcription, translation and subtitle timing into one browser interface, with dubbing as an optional output. The project is open source under Apache 2.0; a separate hosted service offers subtitle translation and dubbing.

RVC WebUI is a local AI voice conversion tool for people who want to train a custom voice, change the voice in a recording, or use a live voice changer. It runs on Windows and Linux, including Ubuntu servers, with a browser interface for training and conversion and a separate interface for live use. It's free and open source under the MIT license.

AllTalk TTS generates speech on your own computer. The project recommends v2 for most users; the saved documentation below describes v1, built on Coqui TTS and XTTSv2 models. It's for people adding voices to AI conversations or producing spoken audio from longer texts. It runs as a standalone application or alongside Text-generation-webui, with support for Windows, Linux and macOS.

WhisperWriter turns microphone speech into text and types it into the window you're working in. It's for people who want voice input in their existing desktop apps, with a choice between transcription on their own computer and an external service. The Python app runs on Windows, macOS and Linux and uses the GPL-3.0 open-source license.

llama.rn brings llama.cpp into React Native apps so developers can run local LLM inference on iOS and Android. It's an MIT-licensed library for building AI features into a mobile app, with model processing on the device. It uses GGUF models and requires React Native's New Architecture.

Scriberr is a free, open source transcription app for people who want to keep meeting recordings and voice notes on their own hardware. It turns audio and video into text locally, with offline transcription once the models are downloaded. The MIT license allows you to use and modify it.

Witsy is an AGPL-3.0 desktop AI assistant for macOS, Windows and Linux that connects MCP tools to local and cloud models. It's for people who want document chat, writing help and voice features in one app, with a choice of where their models run.

Wyoming Piper connects Piper's local text-to-speech engine to Home Assistant and other clients that use the Wyoming protocol. It's for people building a voice assistant on their own hardware who need speech generation as a self-hosted service. The project is open source under the MIT license.

Magenta RealTime 2 is a local AI music model and synthesis engine for musicians and developers who want to play or build AI musical instruments on a laptop. It generates streaming audio in real time, with open weights and code under the Apache 2.0 license.

Enconvo is a native AI assistant and agent for Mac that can use your screen and selected text as context, then work inside your apps. It's for people who want help with writing, research and everyday tasks alongside the app they're using. A sidebar keeps the agent beside the current app, while text selection tools give quick access to editing and translation.

TTS WebUI brings local text-to-speech, music generation and audio processing into one browser interface. It's for people creating spoken audio or music, and for developers who want to add speech to a self-hosted chat app. The interface combines Gradio and React, with extensions that let you choose which audio models to use.

Home-LLM connects Home Assistant to language models running on your own hardware, so you can control smart devices through voice or chat. It's for Home Assistant users who want natural language control without relying on a cloud service or subscription. The project pairs a custom integration with small models trained specifically for smart home commands.

GPT-SoVITS is a local text-to-speech and voice cloning tool. It can generate speech from a short reference recording or fine-tune a model for a custom voice. The source code uses the MIT license.

OpenVoice is an open-source voice cloning tool that uses a short recording to reproduce a speaker's voice in generated speech. It's for developers and creators who need a recognizable voice across languages, with control over how that voice sounds. The Python project is MIT licensed for commercial use.

Voice Changer (w-okada), also called VCClient, converts your voice as you speak using AI voice models. It's for people who want live voice conversion on their own computer, including those recording gaming commentary while running demanding software. Processing can stay local.

OpenVoiceOS is a free, open-source voice AI platform for people building their own smart speakers or adding voice control to devices. Its core builds on a fork of MycroftAI/mycroft-core, and most classic Mycroft skills also work with it. The project uses the Apache 2.0 license, which permits personal and commercial use.

Wyoming Protocol connects Home Assistant with separate voice services on a trusted network. It's for developers and people assembling a self-hosted voice assistant who want to choose their own speech recognition, speech synthesis and wake word components. The Python project is open source under the MIT license.

LiveKit Agents is a framework for developers building voice assistants, phone agents, and apps that combine speech with video or text. Agents join LiveKit rooms as participants, so they can interact with people through web and mobile apps or telephone calls. The Apache 2.0 project lets you run the entire stack on your own servers, including the LiveKit media server.

Rasa is an AI agent platform for product teams building customer-facing text and voice assistants. Teams can deploy agents on their own infrastructure and choose their models and data arrangements. Its CALM engine combines language model understanding with business flows whose code enforces rules, so an assistant can handle conversational wording while following defined processes.

Parakeet is NVIDIA's speech recognition model family. The linked parakeet-tdt-0.6b-v2 is its English speech-to-text model for developers and researchers building transcription services, subtitles or voice applications. It runs locally through NeMo on Linux, with NVIDIA GPUs recommended for inference. It's a model you can embed in an application, rather than a desktop transcription app.

Distil-Whisper is a family of local speech recognition models for developers building English transcription into their apps or services. It reduces Whisper's size and processing time while retaining much of its transcription accuracy. It supports English only.

BoltAI is a native Mac app for people who want local LLMs and cloud AI services in the same workspace. It runs on Intel and Apple Silicon Macs and suits coding, writing, research and document analysis. The interface uses SwiftUI and AppKit.

Moshi is a voice AI model and dialogue framework that can listen while it speaks. It processes speech directly, retaining information such as emotion and non-verbal cues that a text transcription can miss. It's aimed at researchers and developers building spoken AI applications, with local inference and self-hosted server options.

nerd-dictation is an open-source dictation utility for desktop Linux that recognizes speech locally through VOSK. It's for people who want voice input in their existing applications and are comfortable with a command-line tool. Audio processing stays on your machine, and recognition works offline.

OpenSuperWhisper is a local speech-to-text app for people who want to dictate or transcribe recordings on an Apple Silicon Mac. It supports Whisper and Parakeet, with model downloads available inside the app. The project is open source under the MIT license.

TEN Framework is a self-hosted framework for developers building voice AI agents and multimodal conversational apps. It focuses on low-latency, real-time conversations and supports both RTC and WebSocket connections. You can run its agent examples locally with Docker or deploy them on your own server.

Whisper WebUI turns audio into transcripts and subtitles through a browser interface that runs on your own machine or a self-hosted server. It's for people captioning videos, transcribing recordings or translating spoken content who want local speech processing. The project is open source under Apache 2.0 and supports Docker and Pinokio.

Mattermost Agents brings AI assistants into a self-hosted Mattermost workspace, with a choice of local models or cloud providers. It's for teams that want help with long discussions and meeting recordings while controlling where their collaboration data goes. The plugin is open source under Apache 2.0.

Leon is a self-hosted personal AI assistant for Linux, macOS and Windows. The project focuses on the 2.0 Developer Preview on its develop branch. Its MIT-licensed runtime combines tools, context and memory with a local web interface.

Willow is a self-hosted voice assistant platform for people who want home automation voice control on their own hardware. It runs on Espressif's ESP32-S3-BOX family and connects to Home Assistant, openHAB, or other services that accept speech results over HTTP. The project is open source under Apache 2.0.