Open Source Models for Voice AI

Dograh — Open-source voice platform for custom, self-hosted voice agents and pipelines.

FireRedTeam — FireRedASR is FireRedTeam's open-source automatic speech recognition (ASR) model for high-accuracy speech-to-text.

Resemble AI — SoTA open-source TTS

TEN Framework — Voice Activity Detector (VAD) : low-latency, high-performance and lightweight

Whisper — Robust Speech Recognition via Large-Scale Weak Supervision

Vosk — Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node

LiveKit — Open source platform for building, testing, deploying, and scaling voice, video, and physical AI agents.

VoiceMesh — A production-inspired reliability lab for real-time voice AI infrastructure — full VAD to STT to LLM to TTS pipeline with observability, event streaming, durable workflows, and billing.

Kaldi — kaldi-asr/kaldi is the official location of the Kaldi project.

Gemma — Gemma open-weight LLM library, from Google DeepMind

Agora — Real-time voice AI platform enabling natural, low-latency conversations.

MiniMax — MiniMax-M3 is an open-source large language model released by MiniMax, available on GitHub to run and build on.

Gradium — Build expressive, low latency voice AI with a unified API for text, speech, and cloning.

Moonshot AI — MoBA: Mixture of Block Attention for Long-Context LLMs

QwenLM — Qwen3-ASR is an open-source series of ASR models developed by the Qwen team at Alibaba Cloud, supporting stable multilingual speech/music/song recognition, language detection and timestamp prediction.

pyannote.audio — Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding

Dia2-2B — Dia2-2B by nari-labs — open-source model on Hugging Face.

FireRedTeam — FireRedChat is FireRedTeam's open-source real-time, full-duplex conversational voice agent.

OpenHome — OpenHome — open-source OS for AI voice agents, LLM-agnostic and ship-ready for real hardware.

DeepSeek-V3 — DeepSeek-V3 — open source.

Kyutai — Moshi is a speech-text foundation model and full-duplex spoken-dialogue framework, using Mimi, a state-of-the-art streaming neural audio codec — enabling real-time, natural voice conversation.

Izwi — Voice AI runtime for private deployment with an OpenAI-compatible API.

Myshell — High-quality multi-lingual text-to-speech library by MyShell.ai. Support English, Spanish, French, Chinese, Japanese and Korean.

Voicebox — Open Source Voice Cloning Desktop App for Mac, Windows, and Linux.

Llama — Inference code for Llama models

F5-TTS — Official code for "F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching"

QwenLM — Qwen3.6 is the large language model series developed by Qwen team, Alibaba Group.

Real-Time Voice Cloning — Clone a voice in 5 seconds to generate arbitrary speech in real-time

NVIDIA NeMo — A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech)

Kyutai — Kyutai's open speech-to-text and text-to-speech models built on the Delayed Streams Modeling framework for streaming, low-latency speech.