funasr
OpenAI-compatible speech recognition toolkit with WebSocket streaming, vLLM acceleration, and llama.cpp/GGUF edge runtime.
Decision gist · record as of 2026-08-14
Yes. FunASR is actively maintained, MIT-licensed, and offers a genuine alternative to single-model approaches. Install it if you need speaker diarization, emotion detection, streaming support, or self-hosted deployment. The 26 runtime dependencies add setup complexity, but the toolkit's flexibility across models and deployment targets justifies it for production use. Start with the Colab quickstart to test before local setup.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires Python ≥ 3.7.
- Dependencies include transformers, librosa, scipy, and others; GPU-accelerated models require additional setup.
- CPU-only inference is viable but slower than GPU deployment.
License · maintenance · safety
The MIT License (permissive) — MIT license (permissive) allows commercial and private use without restriction. Toolkit itself is free and self-hosted; individual model licenses vary by checkpoint.
last release 2026-08-14 (0 days) · last repo commit 2026-08-14 · 19,836 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 497,863 downloads/mo, #6,331 on PyPI
Alternatives
Verify before relying
pip install funasr
from funasr import AutoModel
model = AutoModel(model="iic/SenseVoiceSmall", device="cpu")
result = model.generate(input="audio.wav")
print(result[0]["sentence_info"])- Whether all 26 runtime dependencies are required for basic usage or only for specific models/features.
- Specific performance metrics and whether they apply to typical hardware configurations.
- Detailed compatibility matrix between model checkpoints and deployment targets (CPU, GPU, GGUF, vLLM).
- Whether external PyTorch installation is a hard blocker or handled by dependency resolution.
What it is and what it does
FunASR is a production-grade speech recognition toolkit that goes beyond single-model transcription. It coordinates multiple specialized models—ASR, voice activity detection (VAD), speaker diarization, punctuation, and emotion tagging—into unified pipelines. You choose the checkpoint that fits your task: Fun-ASR-Nano for Chinese/English/Japanese with GPU acceleration, SenseVoiceSmall for CPU-viable five-language ASR with emotion detection, Paraformer for low-latency streaming, or Qwen3-ASR for 52-language coverage. The toolkit handles offline batch processing, real-time WebSocket streaming, and edge deployment via GGUF or vLLM acceleration.
It ships with OpenAI-compatible API serving, MCP server integration, and agent hooks for LangChain and AutoGen. The core dependency stack includes transformers, librosa, scipy, and Hydra for configuration; model weights are fetched from ModelScope or Hugging Face on first use. Unlike single-model alternatives, FunASR lets you trade off speed, language coverage, and feature richness per deployment scenario—from CPU-only inference to GPU-accelerated batch processing.
Use it for
- Transcribe multilingual audio with speaker labels and timestamps for meeting notes, interview processing, or call center analytics.
- Deploy a self-hosted OpenAI-compatible speech API endpoint for voice agents and chatbots without cloud vendor lock-in.
- Extract emotion and audio events alongside transcription for customer sentiment analysis or accessibility.
- Stream live audio transcription via WebSocket for real-time captioning or voice command interfaces.
- Run low-latency ASR on CPU or edge devices using SenseVoiceSmall or GGUF checkpoints for embedded systems.
- Batch-process large audio corpora with vLLM acceleration for industrial data labeling or training dataset preparation.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes.
FunASR is actively maintained, MIT-licensed, and offers a genuine alternative to single-model approaches. Install it if you need speaker diarization, emotion detection, streaming support, or self-hosted deployment. The 26 runtime dependencies add setup complexity, but the toolkit's flexibility across models and deployment targets justifies it for production use. Start with the Colab quickstart to test before local setup.
Install
funasr on PyPI
Before you install
Low friction installation with a pure-Python wheel. Active maintenance (latest release 2026-08-14) and strong community signal (19836 GitHub stars). Requires installation of dependencies including transformers and librosa before use; GPU support is optional but recommended for flagship models.
Requires Python ≥ 3.7. Dependencies include transformers, librosa, scipy, and others; GPU-accelerated models require additional setup. CPU-only inference is viable but slower than GPU deployment.
License in practice
MIT license (permissive) allows commercial and private use without restriction. Toolkit itself is free and self-hosted; individual model licenses vary by checkpoint.
Quickstart
pip install funasr
from funasr import AutoModel
model = AutoModel(model="iic/SenseVoiceSmall", device="cpu")
result = model.generate(input="audio.wav")
print(result[0]["sentence_info"])Verify before relying
- Whether all 26 runtime dependencies are required for basic usage or only for specific models/features.
- Specific performance metrics and whether they apply to typical hardware configurations.
- Detailed compatibility matrix between model checkpoints and deployment targets (CPU, GPU, GGUF, vLLM).
- Whether external PyTorch installation is a hard blocker or handled by dependency resolution.
Package facts
| License | The MIT License permissive |
| Python support | Supports the current Python release >=3.7.0 |
| Install friction | Low. Pure-Python wheel |
| Runtime dependencies | 26 packagesscipylibrosasoundfilenumpyPyYAMLtqdmrequestsregexwebsocketsomegaconfhydra-coremodelscopehuggingface_hubsafetensorstransformerstiktokensentencepiecekaldiiojiebajamojaconvumap_learnrapidfuzztorch_complextensorboardXoss2 |
| Maintenance | Actively maintained 0 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 497,863 / month, #6,331 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Development Status :: 5 - Production/StableIntended Audience :: DevelopersIntended Audience :: Science/ResearchLicense :: OSI Approved :: MIT LicenseOperating System :: MacOSOperating System :: Microsoft :: WindowsOperating System :: POSIX :: LinuxProgramming Language :: PythonProgramming Language :: Python :: 3Programming Language :: Python :: 3.10Programming Language :: Python :: 3.11Programming Language :: Python :: 3.12Programming Language :: Python :: 3.8Programming Language :: Python :: 3.9Topic :: Multimedia :: Sound/Audio :: SpeechTopic :: Scientific/Engineering :: Artificial IntelligenceTopic :: Software Development :: Libraries :: Python Modules |
Evidence: funasr-1.4.2-py3-none-any.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “ASR with speaker diarization”
- funasrFunASR is a speech recognition toolkit that transcribes audio offline…
- whisperxWhisperX performs fast automatic speech recognition with word-level…
- sherpa-onnxSherpa-onnx runs speech recognition, text-to-speech, speaker…
Give your agent the search over MCP, or paste the wish link into any chat.
More Python Modules packages
Converts domain names between Unicode and ASCII-compatible encoding (Punycode) according to IDNA 2008 and Unicode Technical Standard 46, with security validation and broader script coverage than the standard library.
Install it if you work with internationalized domain names, need to validate domains, or use HTTP clients that depend on it transitively.
Setuptools is a Python build backend and package management tool that handles building, distributing, and installing Python packages, including support for C/C++ extension modules.
PyYAML parses and emits YAML 1.1 data format, enabling serialization and deserialization of configuration files and Python objects to and from human-readable YAML text.
Pydantic validates Python data structures against type hints, coercing and checking input at runtime to ensure it matches a declared schema.
Provides reusable metadata objects for use with PEP-593 `typing.Annotated` to express common constraints like bounds, collection sizes, and predicates on types.
Install it if you use or build libraries that need to express type constraints in a standardized, inspectable way—or if you want to annotate your own types with…
Provides runtime tools to inspect and introspect Python type annotations, enabling programmatic examination of type hints at execution time.
See also sherpa-onnx · sherpa-onnx-core · whisper-timestamped · onnx-asr · openai-whisper · SpeechRecognition · qwen-asr · whisperx · livekit-plugins-speechmatics · kaldifst