speech recognition / speech processing toolkit
FunASR is an industrial speech recognition toolkit covering ASR, streaming ASR, VAD, punctuation, speaker diarization, and OpenAI-compatible serving, with pre-trained models for multiple languages.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
choose when you need NVIDIA's NeMo ecosystem with multi-modal speech+LLM training and deployment.
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)
choose when you are in the PaddlePaddle ecosystem and want a full ASR/TTS toolkit.
Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text frontend, Speaker Verification System, End-to-End Speech Translation and Keyword Spotting. Won NAACL2022 Best Demo Award.
choose for a research-friendly PyTorch toolkit with many recipes for speech and beyond.
A PyTorch-based Speech Toolkit
choose for a research-oriented end-to-end speech toolkit with extensive recipe support.
End-to-End Speech Processing Toolkit
choose for a production-first streaming ASR toolkit with training and lightweight runtime.
Production First and Production Ready End-to-End Speech Recognition Toolkit
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
choose for a specialized low-latency voice-agent ASR model with intent recognition.
Very low latency speech to text, intent recognition, and text to speech, for building voice agents and interfaces
choose for a drop-in real-time STT library with VAD and wake word, based on Whisper.
A robust, efficient, low-latency speech-to-text library with advanced voice activity detection, wake word activation and instant transcription.
choose when you only need a strong multilingual ASR model with emotion/event detection.
Open-source SenseVoiceSmall model for Mandarin, Cantonese, English, Japanese, and Korean ASR, language ID, emotion recognition, and audio event detection.
choose for Apple-native speech models (ASR/TTS/VAD) via CoreML in Swift.
Frontier CoreML audio models in your apps — text-to-speech, speech-to-text, voice activity detection, and speaker diarization. In Swift, powered by SOTA open source.
choose for Apple Silicon apps wanting MLX/CoreML speech models in Swift.
AI speech toolkit for Apple Silicon — ASR, TTS, speech-to-speech, VAD, and diarization powered by MLX and CoreML
choose for a simple Whisper+pyannote pipeline if you need ASR plus diarization in a notebook.
Automatic Speech Recognition with Speaker Diarization based on OpenAI Whisper
choose for a ready-made real-time local ASR server with OpenAI-compatible APIs, based on Whisper.
Real-time, local speech-to-text with streaming ASR, speaker diarization, translation, and OpenAI/Deepgram-compatible APIs.
choose for high-performance C++ Whisper inference without Python.
High-performance GPGPU inference of OpenAI's Whisper automatic speech recognition (ASR) model
choose for accurate word-level timestamps and diarization on top of Whisper in Python.
WhisperX: Automatic Speech Recognition with Word-level Timestamps (& Diarization)
choose for lightweight word-level timestamps from Whisper with confidence scores.
Multilingual Automatic Speech Recognition with word-level timestamps and confidence
choose for on-device real-time ASR/VAD with ncnn on many platforms.
Real-time speech recognition and voice activity detection (VAD) using next-gen Kaldi with ncnn without Internet connection. Support iOS, Android, Linux, macOS, Windows, Raspberry Pi, VisionFive2, LicheePi4A etc.
choose if you need an archived all-in-one NLP/speech research platform (no maintenance).
DELTA is a deep learning based natural language and speech processing platform. LF AI & DATA Projects: https://lfaidata.foundation/projects/delta/
choose for lightweight offline ASR with simple APIs for mobile and server.
Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node
Does the same job, but ships as an app. Useful to a person, not swappable into a codebase.
choose for a quick REST wrapper around Whisper instead of building your own server.
OpenAI Whisper ASR Webservice API
choose to run Whisper on Windows as an executable without installing Python.
Whisper & Faster-Whisper standalone executables for those who don't want to bother with Python.
These projects were analysed and named FunASR among their alternatives. The relationship is not symmetric — how FunASR rates them is a separate judgement, made when FunASR is analysed in its own right.
calls FunASR “choose when you need a deployment-focused ASR toolkit with built-in serving and MIT-licensed stack, less GPU-heavy than 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)
calls FunASR “when you need a trainable ASR toolkit with VAD/punctuation and OpenAI-compatible serving, especially for Chinese”
Real-time, local speech-to-text with streaming ASR, speaker diarization, translation, and OpenAI/Deepgram-compatible APIs.
calls FunASR “Choose when you need a production ASR toolkit with streaming, VAD, punctuation, and OpenAI-compatible/MCP serving.”
Robust Speech Recognition via Large-Scale Weak Supervision
calls FunASR “Choose for a production-grade ASR toolkit with streaming, VAD, and diarization, if Python/ModelScope deployment is acceptable.”
Port of OpenAI's Whisper model in C/C++
calls FunASR “use it for a modular ASR toolkit with streaming, VAD, punctuation, diarization and training, not one unified pretrained model”
Open-Source Frontier Voice AI
calls FunASR “Choose if you need a full ASR toolkit with streaming, VAD, punctuation, and diarization alongside a different model family.”
Faster Whisper transcription with CTranslate2
calls FunASR “full ASR toolkit with training and serving, more flexible but far heavier for simple transcription.”
Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node
calls FunASR “Use when you want a Python ASR toolkit with industrial-grade streaming ASR, VAD, punctuation, and OpenAI-compatible/MCP serving on servers, not embedded devices.”
Speech-to-text, text-to-speech, speaker diarization, speech enhancement, source separation, and VAD using next-gen Kaldi with onnxruntime without Internet connection. Support embedded systems, Android, iOS, HarmonyOS, Raspberry Pi, RISC-V, RK NPU, Axera NPU, Ascend NPU, x86_64 servers, websocket server/client, support 12 programming languages
calls FunASR “choose when you need a production ASR pipeline with streaming, VAD, and OpenAI-compatible serving, and you can run server-side.”
Very low latency speech to text, intent recognition, and text to speech, for building voice agents and interfaces
calls FunASR “Full-featured ASR toolkit with training/streaming; pick it when you need an all-in-one pipeline in Python rather than low-level Whisper inference.”
High-performance GPGPU inference of OpenAI's Whisper automatic speech recognition (ASR) model
calls FunASR “choose when you need ASR and diarization integrated in one production-ready toolkit.”
Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding
calls FunASR “Choose when you need training/finetuning or an all-in-one ASR/VAD/diarization pipeline; it is much heavier than RealtimeSTT.”
A robust, efficient, low-latency speech-to-text library with advanced voice activity detection, wake word activation and instant transcription.
calls FunASR “choose for a full ASR toolkit with training, streaming, and pipelines.”
Speech recognition module for Python, supporting several engines and APIs, online and offline.
calls FunASR “ASR-focused toolkit with streaming, VAD, punctuation and diarization; choose for production ASR pipelines.”
Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text frontend, Speaker Verification System, End-to-End Speech Translation and Keyword Spotting. Won NAACL2022 Best Demo Award.
calls FunASR “Choose when your focus is ASR and you want streaming, VAD, punctuation, and scalable serving all in one box.”
A PyTorch-based Speech Toolkit
calls FunASR “Choose when you want a focused ASR pipeline with VAD, punctuation, and serving, but don't need TTS or other ESPnet features.”
End-to-End Speech Processing Toolkit
calls FunASR “Choose when you need only speech recognition with streaming/VAD/punctuation and prefer a non-MLX toolkit.”
A text-to-speech (TTS), speech-to-text (STT) and speech-to-speech (STS) library built on Apple's MLX framework, providing efficient speech analysis on Apple Silicon.