sherpa-onnx Releases

This update enhances cross-platform compatibility and stability, adding support for newer CUDA versions on ARM64, fixing Android TTS crashes, and improving build scripts.

Mar 21, 2025

This release exports the Vocos vocoder to the sherpa-onnx toolkit and provides a C++ runtime for its usage.

Mar 17, 2025

This major version introduces RKNN support for Zipformer CTC and Ebranchformer models, exports GTCRN speech enhancement models, and adds extensive multi-language APIs for speech enhancement.

Mar 16, 2025

This release documents and provides model details for GTCRN and DPDFNet speech enhancement models, including parameter counts and intended use cases for various hardware tiers.

Mar 10, 2025

This release focuses on stability fixes, including JNI exception handling, Whisper token normalization, and various build/publishing improvements for different platforms and architectures.

Feb 26, 2025

This release adds support for Sonnet 3.7 and Perplexity Deep Research, introduces a 'Thinking' option for LLM text generation, and updates DeepSeek R1 implementation.

Feb 26, 2025

This release exports the FireRedASR model to ONNX and provides APIs for it across multiple programming languages.

Feb 17, 2025

This release updates the project branding to 'TypedAI' and includes UI tweaks, GitLab code review agent improvements, and fixes for various LLM agents and integrations.

Feb 17, 2025

Feb 13, 2025

Enhances Kokoro TTS with an MFC example and fixes Windows gb2312 encoding issues, while also improving Linux aarch64 build support.

Feb 9, 2025

A major update introducing comprehensive Kokoro TTS 1.0 support across multiple platforms and programming languages, alongside HarmonyOS keyword spotting capabilities.

Feb 7, 2025

Adds support for reasoning models (o1, o3, DeepSeek R1) and features like multi-agent debates and UI enhancements.

Feb 3, 2025

Adds iOS examples for Matcha and Kokoro TTS, fixes Android TTS UI and keyword spotting issues, and continues expanding HarmonyOS support.

Jan 22, 2025

Introduces the Kokoro TTS model with full API support across C++, Python, C#, Swift, Go, Dart, Pascal, JavaScript, and Android.

Jan 17, 2025

Resolves build issues, particularly when TTS components are disabled, and fixes tensor device handling in export scripts.

Jan 13, 2025

Focuses on Matcha-TTS support with new APIs, espeak-ng integration, and HarmonyOS examples, alongside a .NET 8 upgrade.

Jan 6, 2025

Updates AI SDK and adds support for new models (Gemini 2.0, DeepSeek v3) and features like PDF attachments.

Jan 6, 2025

Adds new TTS models for Latvia and Persian, introduces a byte-level BPE Zipformer tokenizer, and provides C++ runtime for Matcha-TTS.

Dec 31, 2024

Minimal release notes: "The hifigan vocoder models are exported from https://drive.google.com/drive/folders/1-eEYTB5Av9jNql0WGBlRoi-WH2J7bp5Y"

Dec 30, 2024

This release focuses on Android enhancements, including static linking for ONNXRuntime and support for byte-level BPE models. It also introduces background threading for VAD processing and updates ONNX Runtime to version 1.16.0 for Jetson devices.

Dec 24, 2024