sherpa-onnx
Decision gist · record as of 2026-08-14
Yes. Sherpa-onnx is actively maintained, permissively licensed, has no known vulnerabilities, and offers broad platform and task coverage for local audio AI. Install it if you need on-device speech or audio processing without external dependencies. The medium install friction is offset by prebuilt wheels and strong community adoption.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires a pre-trained ONNX model file (not included in the package); audio input must be in a supported format.
- Medium install friction due to compiled wheels for multiple Python versions and architectures; however, prebuilt wheels are available for common platforms (Linux x86_64, macOS, Windows, ARM variants).
- Active maintenance with recent releases (3 days since last update) and 14181 repository stars suggest reliable ongoing support.
License · maintenance · safety
permissive license (permissive) — Apache licensed under permissive terms, allowing commercial and private use with minimal restrictions; suitable for most production and proprietary projects.
last release 2026-08-11 (3 days) · last repo commit 2026-08-13 · 14,181 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 634,735 downloads/mo, #5,642 on PyPI
Alternatives
Verify before relying
pip install sherpa-onnx
import sherpa_onnx
# Requires a pre-trained ONNX model file and audio input
recognizer = sherpa_onnx.OfflineRecognizer.from_pretrained(...)
result = recognizer.recognize(audio_data)- Whether pre-trained models are bundled or must be downloaded separately
- Performance characteristics (latency, memory usage) on different hardware
- Supported audio formats and sample rates for each task
- Whether GPU acceleration is available or only CPU inference
What it is and what it does
Sherpa-onnx is a Python wrapper around ONNX Runtime for running speech and audio AI models locally without cloud dependencies. It supports a wide range of tasks—speech-to-text (streaming and offline), text-to-speech, speaker identification and diarization, voice activity detection, keyword spotting, audio tagging, speech enhancement, and source separation—across diverse platforms including Linux, macOS, Windows, Android, iOS, and specialized hardware like Raspberry Pi and NVIDIA Jetson boards.
The package is built on ONNX Runtime and requires sherpa-onnx-core as its runtime dependency. It targets developers who need on-device audio AI without external API calls, with support for multiple programming languages and NPU accelerators (Rockchip, Qualcomm, Ascend). Installation uses prebuilt wheels for Python 3.7+ on common architectures, though model files must be obtained separately.
Use it for
- Build offline speech-to-text applications for edge devices or privacy-sensitive environments
- Add real-time voice commands and keyword spotting to embedded systems or IoT devices
- Implement speaker identification or diarization for audio analysis and meeting transcription
- Deploy text-to-speech synthesis on mobile or server applications without cloud API costs
- Process audio locally on Raspberry Pi, Jetson, or other ARM-based platforms
- Integrate voice activity detection or speech enhancement into audio pipelines
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes.
Sherpa-onnx is actively maintained, permissively licensed, has no known vulnerabilities, and offers broad platform and task coverage for local audio AI. Install it if you need on-device speech or audio processing without external dependencies. The medium install friction is offset by prebuilt wheels and strong community adoption.
Install
sherpa-onnx on PyPI
Before you install
Medium install friction due to compiled wheels for multiple Python versions and architectures; however, prebuilt wheels are available for common platforms (Linux x86_64, macOS, Windows, ARM variants). Active maintenance with recent releases (3 days since last update) and 14181 repository stars suggest reliable ongoing support.
Requires a pre-trained ONNX model file (not included in the package); audio input must be in a supported format.
License in practice
Apache licensed under permissive terms, allowing commercial and private use with minimal restrictions; suitable for most production and proprietary projects.
Quickstart
pip install sherpa-onnx
import sherpa_onnx
# Requires a pre-trained ONNX model file and audio input
recognizer = sherpa_onnx.OfflineRecognizer.from_pretrained(...)
result = recognizer.recognize(audio_data)Verify before relying
- Whether pre-trained models are bundled or must be downloaded separately
- Performance characteristics (latency, memory usage) on different hardware
- Supported audio formats and sample rates for each task
- Whether GPU acceleration is available or only CPU inference
Package facts
| License | permissive license permissive |
| Python support | Supports the current Python release >=3.7 |
| Install friction | Medium. Platform-specific wheel |
| Runtime dependencies | 1 packagesherpa-onnx-core |
| Maintenance | Actively maintained 3 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 634,735 / month, #5,642 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Programming Language :: C++Programming Language :: PythonTopic :: Scientific/Engineering :: Artificial Intelligence |
Evidence: sherpa_onnx-1.13.5-cp310-cp310-linux_armv7l.whl; sherpa_onnx-1.13.5-cp310-cp310-macosx_10_15_universal2.whl; sherpa_onnx-1.13.5-cp310-cp310-macosx_10_15_x86_64.whl; sherpa_onnx-1.13.5-cp310-cp310-macosx_11_0_arm64.whl; sherpa_onnx-1.13.5-cp310-cp310-manylinux2014_aarch64.manylinux_2_17_aarch64.whl; sherpa_onnx-1.13.5-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl; sherpa_onnx-1.13.5-cp310-cp310-win32.whl; sherpa_onnx-1.13.5-cp310-cp310-win_amd64.whl; sherpa_onnx-1.13.5-cp311-cp311-linux_armv7l.whl; sherpa_onnx-1.13.5-cp311-cp311-macosx_10_15_universal2.whl; sherpa_onnx-1.13.5-cp311-cp311-macosx_10_15_x86_64.whl; sherpa_onnx-1.13.5-cp311-cp311-macosx_11_0_arm64.whl; sherpa_onnx-1.13.5-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.whl; sherpa_onnx-1.13.5-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.whl; sherpa_onnx-1.13.5-cp311-cp311-win32.whl; sherpa_onnx-1.13.5-cp311-cp311-win_amd64.whl; sherpa_onnx-1.13.5-cp311-cp311-win_arm64.whl; sherpa_onnx-1.13.5-cp312-cp312-linux_armv7l.whl; sherpa_onnx-1.13.5-cp312-cp312-macosx_10_15_universal2.whl; sherpa_onnx-1.13.5-cp312-cp312-macosx_10_15_x86_64.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 › “keyword spotting audio”
- sherpa-onnxSherpa-onnx runs speech recognition, text-to-speech, speaker…
- sherpa-onnx-coresherpa-onnx-core provides pre-built ONNX runtime binaries for speech…
- pocketsphinxPocketSphinx provides Python bindings for Carnegie Mellon…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also funasr · onnx-asr · sherpa-onnx-core · silero-vad · pocketsphinx · SpeechRecognition · vosk · qwen-asr · wyoming · pvporcupine