sherpa_onnx

What's Changed

• Fix building android and java examples by @csukuangfj in #3866
• Fix using SPM in iOS examples by @csukuangfj in #3867
• Fix releasing dart packages by @csukuangfj in #3874
• Unify sherpa-onnx init API for Flutter and Dart CLI by @csukuangfj in #3875
• Expose the pyannote window shift ratio in the C API and language bindings by @JulianPscheid in #3870
• Add VAD+ASR Flutter examples by @csukuangfj in #3877
• Release v1.13.6 by @csukuangfj in #3879

Full Changelog: v1.13.5...v1.13.6

• Add VAD+ASR Flutter examples with 9 ASR models (#3877)
• Expose the pyannote window shift ratio in the C API and language bindings (#3870)
• Unify sherpa-onnx init API for Flutter and Dart CLI (#3875)
• Fix releasing dart packages (#3874)
• Fix using SPM in iOS examples (#3867)
• Fix building android and java examples (#3866)

What's Changed

• Re-export nemotron-3.5-asr-streaming-0.6b to ONNX by @csukuangfj in #3732
• Fix building Android KWS demo by @csukuangfj in #3733
• Refactor ONNX export script for nemotron-3.5-asr-streaming-0.6b by @csukuangfj in #3734
• Build a docker image to setup the QNN SDK environment by @csukuangfj in #3736
• Export nemotron-3.5-asr-streaming-0.6b to QNN by @csukuangfj in #3741
• Add Android shared-linking support to sherpa-onnx-sys by @willwade in #3739
• Fix build warnings by @csukuangfj in #3742
• Fix ScaleSilence copying past the pause interval when scale > 1 by @ekenberg in #3744
• Guard ScaleSilence against invalid silence scales by @ekenberg in #3745
• Align the silence scale range between Validate() and ScaleSilence by @ekenberg in #3747
• Export Moonshine ASR models to QNN by @csukuangfj in #3756
• Add C++ runtime for Moonshine ASR with QNN for Qualcomm NPU by @csukuangfj in #3758
• Upload vits-piper-pt_BR-cadu-medium to huggingface by @csukuangfj in #3763
• Fix uploading Piper TTS models by @csukuangfj in #3774
• Fix process abort when espeak-ng rejects the voice (e.g. invalid --kokoro-lang) by @ekenberg in #3749
• Fix dropped diacritics in Supertonic TTS text frontend by @bogdanr in #3750
• Add Fono to 'Projects using sherpa-onnx' by @bogdanr in #3775
• Allocate FireRedASR decoder KV cache by estimated need instead of max… by @Kn0688 in #3777
• Add maven examples for Java API by @csukuangfj in #3783
• Add Java API examples with gradle by @csukuangfj in #3784
• Fix greedy search decoding for Nemo streaming transducer by @csukuangfj in #3785
• Refactor building and testing Java API by @csukuangfj in #3786
• Fix adaptive KV cache for models with hard-coded cache length by @Kn0688 in #3787
• Add ten-vad Rust API example to remove silences from a file by @sportiz91 in #3778
• docs(rust): add FunASR Nano offline ASR example (#3210) by @stantheman0128 in #3772
• Fix use-after-free in async TTS progress callbacks by @kecoco16 in #3781
• Make the Pyannote segmentation window shift configurable by @JulianPscheid in #3769
• Print per-stage timing for offline speaker diarization in debug mode by @JulianPscheid in #3768
• Add Domia to the projects using sherpa-onnx by @kecoco16 in #3793
• Forward SessionConfig.-prefixed provider-config keys to Ort::SessionOptions::AddConfigEntry by @kecoco16 in #3792
• Fix macOS release code signatures by @LauraGPT in #3794
• Raise FireRedASR decode cap from 6 to 8 tokens/sec for fast speech by @Kn0688 in #3796
• Set FireRedASR default max_token_per_second to 8 by @Kn0688 in #3799
• Simplify build.gradle.kts for java api examples. by @csukuangfj in #3788
• Test Java API on Windows arm64 by @csukuangfj in #3801
• fix UB by @x6prl in #3806
• fix(tts): insert pad after BOS in PiperPhonemesToIdsVits (#3721) by @stantheman0128 in #3770
• Export Inflect TTS models to sherpa-onnx by @csukuangfj in #3808
• Add API to get onnxruntime version. by @csukuangfj in #3810
• Add Android demo for Inflect TTS by @csukuangfj in #3811
• Build Python wheels for Windows arm64 by @csukuangfj in #3812
• build(cmake): update spacemit-ort version by @xiaotong-spacemit in #3814
• Support swift package manager (SPM) by @csukuangfj in #3816
• Publish an android-x64 .NET runtime package by @kfrancis in #3818
• Fix non_streaming_server.py crash with --decoding-method modified_beam_search by @slipstr34m in #3819
• Fix process crash when --kokoro-voices is missing or empty (#3748) by @slipstr34m in #3820
• Fix streaming paraformer feature frontend drift from its training config by @slipstr34m in #3821
• Use SPM in iOS examples by @csukuangfj in #3822
• Fix iOS examples using swift package manager by @csukuangfj in #3823
• Add Paraformer Rust API example add rust paraformer example by @nati2405 in #3707
• Add missing files for swift package manager by @csukuangfj in #3828
• Set onnxruntime version to 1.27.1 in SPM by @csukuangfj in #3830
• Refactor building xcframework for macOS and iOS by @csukuangfj in #3831
• Support linking onnxruntime statically in xcframework with shared lib. by @csukuangfj in #3833
• Support using SPM for Flutter on macOS and iOS by @csukuangfj in #3835
• Fix building Python wheels on macOS by @csukuangfj in #3837
• Disable pthread support in WebAssembly. by @csukuangfj in #3836
• Fix Dart speaker diarization input buffer leak by @zhangheng2022 in #3838
• Fix offline Paraformer LFR boundary padding by @LauraGPT in #3839
• docs: add qwen-audio-agent to projects using sherpa-onnx by @x-lixu in #3841
• Add Flutter/Dart Web support via WebAssembly by @csukuangfj in #3842
• Build libs and Python wheels for CUDA 13 by @csukuangfj in #3844
• Add punctuation examples for Flutter by @csukuangfj in #3846
• Add Persian NeMo CTC ASR models (shenava koochik/rizeh/rizeh-pizeh) by @mah92 in #3825
• Add CI for Flutter punctuation examples by @csukuangfj in #3847
• Add emotion_id input to offline VITS TTS by @matiaslin in #3849
• Add support for latest DPDFNet models and offline attenuation limit by @danielr-ceva in #3824
• kotlin-api: validate native ptr after creation by @sigkillsdottir in #3851
• Use onnxruntime 1.27.1 for Linux aarch64 and Android by @yentur in #3850
• fix(asr): stop forcing lowercase in mic and ALSA display (#3621) by @stantheman0128 in #3771
• Add Flutter examples for VAD by @csukuangfj in #3856
• Fix Kotlin Syntax errors by @csukuangfj in #3860
• Update onnxruntime to v1.27.1 by @csukuangfj in #3861
• Fix float32 catastrophic cancellation in NemoNormalizePerFeature by @Eoin-Houstoun in #3857
• Fix float32 catastrophic cancellation in variance computation by @csukuangfj in #3862
• Release v1.13.5 by @csukuangfj in #3863

New Contributors

• @willwade made their first contribution in #3739
• @ekenberg made their first contribution in #3744
• @bogdanr made their first contribution in #3750
• @Kn0688 made their first contribution in #3777
• @sportiz91 made their first contribution in #3778
• @stantheman0128 made their first contribution in #3772
• @kecoco16 made their first contribution in #3781
• @LauraGPT made their first contribution in #3794
• @x6prl made their first contribution in #3806
• @xiaotong-spacemit made their first contribution in #3814
• @kfrancis made their first contribution in #3818
• @slipstr34m made their first contribution in #3819
• @nati2405 made their first contribution in #3707
• @zhangheng2022 made their first contribution in #3838
• @x-lixu made their first contribution in #3841
• @sigkillsdottir made their first contribution in #3851
• @yentur made their first contribution in #3850
• @Eoin-Houstoun made their first contribution in #3857

Full Changelog: v1.13.4...v1.13.5

• Update onnxruntime to v1.27.1 (#3861)
• Fix float32 catastrophic cancellation in variance computation (#3862)
• Fix float32 catastrophic cancellation in NemoNormalizePerFeature (#3857)
• Fix Kotlin syntax errors (#3860)
• Add Flutter/Dart Web support via WebAssembly (#3842)
• Support swift package manager (SPM) for Flutter on macOS and iOS (#3835, #3816, #3822, #3823, #3828, #3830, #3831, #3833)
• Add Flutter examples for VAD (#3856)
• Add punctuation examples for Flutter (#3846, #3847)
• Add support for latest DPDFNet models and offline attenuation limit (#3824)
• Add emotion_id input to offline VITS TTS (#3849)
• Add Persian NeMo CTC ASR models (shenava koochik/rizeh/rizeh-pizeh) (#3825)
• Build libs and Python wheels for CUDA 13 (#3844)
• Use onnxruntime 1.27.1 for Linux aarch64 and Android (#3850)
• Export Inflect TTS models to sherpa-onnx (#3808)
• Add Android demo for Inflect TTS (#3811)
• Add C++ runtime for Moonshine ASR with QNN for Qualcomm NPU (#3758)
• Export Moonshine ASR models to QNN (#3756)
• Export nemotron-3.5-asr-streaming-0.6b to QNN (#3741)
• Re-export nemotron-3.5-asr-streaming-0.6b to ONNX (#3732)
• Build docker image to setup the QNN SDK environment (#3736)
• Add Paraformer Rust API example (#3707)
• Add ten-vad Rust API example to remove silences from a file (#3778)
• Add API to get onnxruntime version (#3810)
• Make FireRedASR max token rate configurable per stream (#3796, #3799)
• Make the Pyannote segmentation window shift configurable (#3769)
• Add Java API examples with gradle (#3783, #3784)
• Refactor building and testing Java API (#3786)
• Publish an android-x64 .NET runtime package (#3818)
• Build Python wheels for Windows arm64 (#3812)
• Test Java API on Windows arm64 (#3801)
• Print per-stage timing for offline speaker diarization in debug mode (#3768)
• Add Android shared-linking support to sherpa-onnx-sys (#3739)
• Refactor building xcframework for macOS and iOS (#3831)
• build(cmake): update spacemit-ort version (#3814)
• Fix streaming paraformer feature frontend to match its training config (#3821)
• Fix offline Paraformer LFR boundary padding (#3839)
• Fix non_streaming_server.py crash when using modified_beam_search (#3819)
• Fix greedy search decoding for NeMo streaming transducer (#3785)
• Fix adaptive KV cache for models with hard-coded cache length (#3787)
• Allocate FireRedASR decoder KV cache by estimated need instead of max (#3777)
• Fix Dart speaker diarization buffer leak (#3838)
• Fix use-after-free in async TTS progress callbacks (#3781)
• fix(tts): insert pad after BOS in PiperPhonemesToIdsVits (#3721) (#3770)
• Validate the voices file for Kokoro TTS models (#3820)
• Fix dropped diacritics in Supertonic TTS text frontend (#3750)
• Do not abort the process when espeak-ng rejects the voice (#3749)
• Fix ScaleSilence copying past the pause interval when scale > 1 (#3744)
• Limit silence scale to [0.01, 2] in ScaleSilence (#3745)
• Align the silence scale range between Validate() and ScaleSilence (#3747)
• Fix macOS release code signatures (#3794)
• Forward SessionConfig.-prefixed provider-config keys to Ort::SessionOptions::AddConfigEntry (#3792)
• kotlin-api: validate native ptr after creation (#3851)
• Initialize the debug member for HomophoneReplacerConfig (#3806)
• fix(asr): stop forcing lowercase in mic and ALSA display (#3621) (#3771)
• Disable pthread support in WebAssembly (#3836)
• Fix building Python wheels on macOS (#3837)
• Fix building Android KWS demo (#3733)
• Fix build warnings (#3742)

What's Changed

• Fix releasing Dart iOS package by @csukuangfj in #3680
• Export Parakeet CTC models to QNN by @csukuangfj in #3686
• Add C++ runtime for Parakeet CTC with QNN. by @csukuangfj in #3688
• Build sdist in split-package mode so its metadata declares sherpa-onnx-core by @Eridanus117 in #3690
• Export Parakeet-TDT-CTC to QNN by @csukuangfj in #3692
• Export the transducer branch of parakeet-tdt-ctc to QNN by @csukuangfj in #3693
• Export Whisper models to QNN by @csukuangfj in #3697
• add project: Add SmartSub application details to README.md by @buxuku in #3698
• Add C++ runtime for Whisper with Qualcomm NPU using QNN. by @csukuangfj in #3699
• Fix building with ninja by @csukuangfj in #3702
• Fix #3171 : 去掉生成绝对路径函数中的窄字符WindowsAPI调用，改用宽字符API和C++17 std::filesystem来避免win平台中文路径问题 by @XiaoBaiCanDo in #3255
• Fix installing espeak-ng-data directory by @csukuangfj in #3708
• Remove spaces right before a punctuation in ASR result by @csukuangfj in #3709
• Fix file related operations on Windows by using wide characters by @csukuangfj in #3710
• Remove leading spaces from ASR results by @csukuangfj in #3711
• Update onnxruntime to 1.27.0 by @csukuangfj in #3718
• Export Parakeet TDT models to QNN by @csukuangfj in #3719
• Add C++ runtime for Parakeet TDT models with QNN by @csukuangfj in #3720
• Export nemotron-speech-streaming-en-0.6b to QNN by @csukuangfj in #3725
• Add C++ runtime for nemotron-speech-streaming-en-0.6b with QNN by @csukuangfj in #3728
• Add ApplyHomophoneReplacer in OfflineRecognizerQwen3ASRImpl by @csukuangfj in #3729
• Support per-stream hotwords in the JavaScript (node-addon) API for non-streaming ASR by @cmurrayis in #3723
• Release v1.13.4 by @csukuangfj in #3730

New Contributors

• @Eridanus117 made their first contribution in #3690
• @buxuku made their first contribution in #3698
• @XiaoBaiCanDo made their first contribution in #3255
• @cmurrayis made their first contribution in #3723

Full Changelog: v1.13.3...v1.13.4

• Add ApplyHomophoneReplacer in OfflineRecognizerQwen3ASRImpl (#3729)
• Add C++ runtime for nemotron-speech-streaming-en-0.6b with QNN (#3728)
• Export nemotron-speech-streaming-en-0.6b to QNN (#3725)
• Add C++ runtime for Parakeet TDT models with QNN (#3720)
• Export Parakeet TDT models to QNN (#3719)
• Update onnxruntime to 1.27.0 (#3718)
• Remove leading spaces from ASR results (#3711)
• Fix file related operations on Windows by using wide characters (#3710)
• Remove spaces right before a punctuation in ASR result (#3709)
• Fix installing espeak-ng-data directory (#3708)
• Fix building with ninja (#3702)
• Add C++ runtime for Whisper with Qualcomm NPU using QNN (#3699)
• Export Whisper models to QNN (#3697)
• Export the transducer branch of parakeet-tdt-ctc to QNN (#3693)
• Export Parakeet-TDT-CTC to QNN (#3692)
• Add C++ runtime for Parakeet CTC with QNN (#3688)
• Export Parakeet CTC models to QNN (#3686)
• Support per-stream hotwords in the JavaScript (node-addon) API (#3723)

What's Changed

• Add JavaScript API example for Supertonic3 TTS by @csukuangfj in #3614
• Update Spacemit toolchain download URL and centralize versioning by @alex-spacemit in #3618
• docs: add React Native sherpa-onnx wrapper by @abretonc7s in #3615
• Add async TTS examples, LinearResampler API, and use node-cpal for microphone by @csukuangfj in #3623
• Improve C/C++ API documentation, examples, and CI by @csukuangfj in #3626
• Add doc for Python API by @csukuangfj in #3627
• Fix heap-buffer-overflow in TopkIndex when topk > size by @suykerbuyk in #3628
• build: allow overriding ONNX Runtime for mobile by @abretonc7s in #3617
• Update VoxSherpa TTS link for stable APK releases by @CodeBySonu95 in #3648
• Add streaming zipformer transducer for Qualcomm NPU with QNN. by @csukuangfj in #3645
• Export streaming zipformer transducer to QNN by @csukuangfj in #3653
• Add Android demo for streaming zipformer transducer ASR with QNN by @csukuangfj in #3654
• Upload X-ASR non-streaming zipformer transducer models by @csukuangfj in #3656
• Fix generating subtitles for funasr nano by @csukuangfj in #3659
• Support non-streaming Zipformer transducer for Qualcomm NPU with QNN. by @csukuangfj in #3655
• Support streaming zipformer transducer ASR for QNN. by @csukuangfj in #3660
• Export X-ASR models to sherpa-onnx by @csukuangfj in #3662
• Add a Whisper example for the Rust API by @WuYujie666 in #3661
• Export X-ASR models to QNN by @csukuangfj in #3666
• Add Android demo with QNN for Streaming X-ASR. by @csukuangfj in #3668
• Add offline CATT diacritization (encoder only) by @matiaslin in #3647
• Add Android QNN demo for non-streaming zipformer with Reazonspeech by @csukuangfj in #3670
• Add OfflineDiacritization JAVA bindings by @matiaslin in #3669
• Add multilingual Nemotron-3.5 streaming ASR support by @JulianPscheid in #3671
• Update iOS to use onnxruntime 1.26.0 by @csukuangfj in #3678
• Release v1.13.3 by @csukuangfj in #3679

New Contributors

• @abretonc7s made their first contribution in #3615
• @WuYujie666 made their first contribution in #3661
• @JulianPscheid made their first contribution in #3671

Full Changelog: v1.13.2...v1.13.3

• Update onnxruntime to 1.26.0 for iOS (#3678)
• Add multilingual Nemotron-3.5 streaming ASR support (#3671)
• Add OfflineDiacritization JAVA bindings (#3669)
• Add Android QNN demo for non-streaming zipformer with Reazonspeech (#3670)
• Add offline CATT diacritization (encoder only) (#3647)
• Add Android demo with QNN for Streaming X-ASR (#3668)
• Export X-ASR models to QNN (#3666)
• Add a Whisper example for the Rust API (#3661)
• Export X-ASR models to sherpa-onnx (#3662)
• Support streaming zipformer transducer ASR for QNN (#3660)
• Support non-streaming Zipformer transducer for Qualcomm NPU with QNN (#3655)
• Fix generating subtitles for funasr nano (#3659)
• Upload X-ASR non-streaming zipformer transducer models (#3656)
• Add Android demo for streaming zipformer transducer ASR with QNN (#3654)
• Begin to export streaming zipformer transducer to QNN (#3653)
• Add streaming zipformer transducer for Qualcomm NPU with QNN (#3645)
• Update VoxSherpa TTS link for stable APK releases (#3648)
• Allow overriding ONNX Runtime for mobile (#3617)
• Fix heap-buffer-overflow in TopkIndex when topk > size (#3628)
• Add doc for Python API (#3627)
• Improve C/C++ API documentation, examples, and CI (#3626)
• Add async TTS examples, LinearResampler API, and use node-cpal for microphone (#3623)
• Add React Native sherpa-onnx wrapper (#3615)
• Update Spacemit toolchain download URL (#3618)
• Add JavaScript API examples for Supertonic3 TTS (#3614)

• Support Supertonic 3 TTS (#3605)
• Release Supertonic3 TTS models (#3609)
• Add Android demo for Supertonic3 TTS (#3612)
• Add KittenTTS v0.8 support (#3591)
• Export KittenTTS models (#3604)
• Generate KittenTTS samples (#3607)
• Export NeMo unified streaming 0.6b models (#3602)
• Add buffered RNNT streaming path for Parakeet Unified (#3575)
• Fix WASM onnxruntime pthread abort() crash (#3599)
• Bump emsdk 4.0.23 (#3582)
• Update onnxruntime package details for spacemit (#3610)

• Fix bugs in NeMo transducer modified beam search (#3589)
• Add bounds checks to prevent SIGSEGV in speaker diarization (#3563)
• Replace dead link with archived link (#3576)
• Fix x64 gpu onnxruntime url (#3578)
• Per-stream use_itn override via Stream::GetOptionInt for SenseVoice (#3565)

• Fix Flutter CI (#3560)
• Fix building Flutter Android APPs (#3559)
• Export nvidia/parakeet-unified-en-0.6b to sherpa-onnx (#3556)
• Update nemotron-speech-streaming-en-0.6b (#3555)
• Expose log probabilities in OfflineRecognizerResult for Go binding (#3553)

• Add more Piper TTS models (#3547)
• Add two Piper Chinese TTS models (#3546)
• Upload the piper sq_AL model (#3541)
• Add Albanian (sq_AL) Piper TTS voice by LanguageWeaver (#3539)
• Add Tauri APP demo for VAD+ASR from a microphone (#3540)
• Improve Tauri VAD+ASR example: settings UI, bug fixes, and RTF display (#3538)
• Fix releasing go packages (#3537)
• Fix publishing dart packages for Android (#3522)
• Avoid passing invalid utf8 str to JNI (#3527)
• Build Tauri desktop APPs for VAD+ASR (#3530)

• Add tauri app example for non-streaming ASR + VAD (#3520)
• Check for nullptr in c/cxx API (#3515)
• Update go.mod and go.sum for the sherpa-onnx-go package. (#3516)
• Fix an offset-by-one error in pyannote speaker diarization. (#3514)
• Set context for thread in Ascend NPU (#3512)
• Enable proxy support for downloader via ureq's proxy-from-env feature in rust-binding build.rs (#3507)
• Add script to download all dependencies for offline use (#3506)
• Update Eigen to v5.0.1 (#3505)
• Fix building Python wheels for Windows (#3504)
• Expose memory-related session options via config (#3503)

• Dotnet Android arm64 targeting (#3485)
• Use file path to initialize onnxruntime session. (#3482)
• update onnxruntime package name and hash for riscv64-spacemit (#3481)
• Fix initializing fp32 Qwen3-ASR models. (#3480)
• Fix Qwen3ASR hotwords handling (#3477)
• Update Qwen3 ASR models (#3476)
• Fix building for vad+asr (#3475)
• Add per-stream language hint for Qwen3-ASR (and tighten scaffold cleanup) (#3472)
• Add hotwords argument to OfflineQwen3ASRModelConfig (#3468)

• Add Go API for Cohere Transcribe (#3466)
• Add Rust API for Cohere Transcribe (#3465)
• Add Dart API for Cohere Transcribe (#3464)
• Add C# API for Cohere Transcribe (#3462)
• Add Pascal API for Cohere Transcribe (#3463)
• Add Java and Kotlin API for Cohere Transcribe (#3461)
• Add JavaScript API for Cohere Transcribe. (#3458)
• Add Swift API for Cohere Transcribe (#3460)
• Add C API and CXX API for Cohere transcribe models (#3457)
• Add C++ runtime and Python API for Cohere Transcribe (#3456)
• Upload models for https://huggingface.co/CohereLabs/cohere-transcribe-03-2026 (#3453)
• Add issue template (#3452)
• Add hotwords support for Qwen3-ASR (#3434)
• Fix links in README.md to use correct casing (#3451)
• Add Wake Word and decibri to projects using sherpa-onnx (#3447)
• Remove num_threads assertion from OnlineRecognizer model configs (#3448)
• Add Speed of Sound to projects using sherpa-onnx (#3438)
• Fix Swift tests (#3433)
• spacemit provider: add provider with config file, then update to 2.0.2 (#3421)
• Add Go API for source separation (#3432)
• Fix C API for reading multi-channel wave files (#3430)
• Add Swift API for source separation (#3426)
• Fix CI tests (#3429)

• Add Pascal API for Qwen3 ASR (#3424)
• Add Rust API for Qwen3 ASR (#3423)
• Add Dart API for Qwen3 ASR (#3422)
• Add Java API and Kotlin API for Qwen3 ASR (#3420)
• Add JavaScript API (node-addon) for Qwen3 ASR (#3419)
• Add JavaScript API (WebAssembly) for Qwen3 ASR (#3416)
• Refactor Go API examples to use sherpa_onnx.ReadWave() (#3415)
• Fix warnings in Go API (#3414)
• Add C# API for Qwen3 ASR (#3413)
• Add Go API for Qwen3 ASR (#3412)
• Update Swift API example for Qwen3 ASR (#3411)
• Update tests for Qwen3 ASR (#3410)
• Upload Qwen3 ASR 0.6B int8 models (#3409)
• Add Qwen3-ASR support (#3399)
• Fix TTS deprecated warnings (#3407)
• Add C# API for source separation (#3406)
• Add CXX API for source separation (#3405)
• Add C API for source separation (#3404)
• Fix building Rust doc (#3403)
• Update Python subtitle script to support FireRedASR CTC and FunASR Nano (#3400)

• Add Rust examples for simulated streaming ASR with VAD (#3398)
• Add Rust example for real-time ASR + VAD (#3397)
• Fix printing microphone sample rate. (#3396)
• Add missing fields of OnlineRecongizerResult to Go API. (#3395)
• Add examples for punctuations (#3394)
• Fix building MFC examples for TTS (#3388)

• Support static link for Rust package. (#3386)
• Test Rust API on Windows (#3385)
• Add VoxSherpa TTS to Projects using sherpa-onnx (#3384)

• Fix building har for OHOS (#3361)
• Refactor MatchaTTS to use the new Generate API (#3362)
• Refactor Kokoro TTS to use the new Generate API (#3363)
• Refactor KittenTTS to use the new Generate API (#3364)
• Refactor VITS to use the new Generate API (#3365)
• Add Rust API examples for TTS (#3366)
• Fix Swift tests (#3367)
• Add Rust API for audio tagging (#3368)
• Add Rust API for speaker embedding extractor and manager (#3369)
• Add Rust API for speaker diarization (#3370)
• Refactor Rust API for speech denoiser (#3371)
• Add Rust API for KWS, offline punctuation and spoken language identification (#3372)
• Add doc for c api and cxx api (#3374)
• Add link to C API doc (#3375)
• Add doc for Rust API (#3376)
• Add doc for Dart API (#3377)
• Add more doc for Rust API (#3378)

• Fix typos in the project (#3293)
• Fix WebAssembly JavaScript API (#3294)
• Remove unnecessary SHERPA_ONNX_API from C/C++ APIs (#3295)
• Fix bugs in CXX APIs (#3296)
• Result goes to stdout (#3274)
• Small fix to online recognizer C++ code (#3297)
• Small fixes to JNI wrappers (#3298)
• Add SetOption/GetOption to OnlineStream and OfflineStream (#3307)
• Add SetOption/GetOption C API and export symbols (#3308)
• Add SetOption/GetOption CXX wrapper for OnlineStream and OfflineStream (#3309)
• Migrate Paraformer is_final to use SetOption mechanism (#3310)
• Add SetOption/GetOption Python bindings for OnlineStream and OfflineStream (#3311)
• Add SetOption/GetOption Java, Kotlin, and JNI bindings (#3312)
• Add SetOption/GetOption Go bindings for OnlineStream and OfflineStream (#3313)
• Add SetOption/GetOption C# bindings for OnlineStream and OfflineStream (#3314)
• Add SetOption/GetOption WASM/JavaScript bindings (#3315)
• Fix padding bug in test-onnx-streaming.py (#3318)
• Fix style issues (#3321)
• Add DPDFNet speech denoiser support for offline and streaming (#3276)
• Upload DPDFNet models (#3322)
• Add C API example for online punctuation (#3323)
• Add online speech denoiser for GTCRN and examples (#3324)
• Add Go API example for online punctuation (#3325)
• Release Rust package for offline/online speech denoiser (#3328)
• Refactor Dart API to check for nullptr (#3329)
• Use onnxruntime v1.23.2 for Android (#3330)
• Refactor ZipVoice TTS to support callback (#3332)
• Add C and CXX API examples for ZipVoice (#3333)
• Add Go API examples for ZipVoice (#3334)
• Add Python API examples for ZipVoice TTS (#3335)
• Add WebAssembly example for ZipVoice (#3337)
• Update WebAssembly download progress text to show MB (#3338)
• Add WebAssembly example for PocketTTS (#3340)
• Add JavaScript (WebAssembly) example for ZipVoice TTS (#3341)
• Add JavaScript (node-addon) example for ZipVoice TTS (#3342)
• Add JavaScript playback examples for Pocket and Supertonic TTS (#3343)
• Add Kotlin and Java API for ZipVoice models (#3344)
• Add C# API examples for ZipVoice models (#3345)
• Add Swift API examples for ZipVoice models (#3346)
• Add Dart API examples for ZipVoice models (#3347)
• Add Rust API example for online punctuation (#3348)
• Add fcitx5-vinput to projects using sherpa-onnx (#3350)
• Add Pascal API examples for ZipVoice models (#3351)
• Add Rust API examples for ZipVoice models (#3352)
• Add SetOption/GetOption/HasOption Kotlin bindings (#3354)
• Fix building Python wheels for Windows (#3355)
• Fix OHOS APIs for TTS and ASR (#3356)
• Add HarmonyOS APIs for online punctuation (#3357)
• Add HarmonyOS APIs for offline punctuation (#3359)

• Add Supertonic TTS support (#3094)
• Upload supertonic tts models (#3263)
• Add Python API examples for Supertonic TTS (#3264)
• Support dynamic decoder layers in canary model runtime (#3268)
• Add CXX API for Supertonic TTS (#3280)
• Add C# API for Supertonic TTS (#3283)
• Add Go API for Supertonic TTS (#3284)
• Add Rust API for Supertonic TTS (#3285)
• Add Swift API for Supertonic TTS (#3286)
• Add JavaScript API for Supertonic TTS (#3287)
• Add Dart API for Supertonic TTS (#3288)
• Add Java and Kotlin API for Supertonic TTS (#3289)
• Add Pascal API and example for Supertonic TTS (#3290)
• Publish pdb files for Debug build on Windows (#3252)
• Fix memory leak in WebAssembly for TTS (#3259)
• Refactor WebAssembly TTS API (#3260)

• Add C++ runtime support for Moonshine v2 (#3232)
• Export Moonshine v2 models to sherpa-onnx (#3234)
• Update Python APIs for Moonshine v2 models (#3235)
• Add Kotlin and Java APIs for Moonshine v2 models (#3237)
• Add C and C++ API for Moonshine v2 models (#3238)
• Add Swift API for Moonshine v2 models (#3240)
• Add JavaScript API (WebAssembly) for Moonshine v2 models (#3241)
• Add JavaScript API (node-addon) for Moonshine v2 models (#3242)
• Add C# API for Moonshine v2 (#3243)
• Add Go API for Moonshine v2 (#3244)
• Add Dart API for Moonshine v2 (#3245)
• Add Rust API for Moonshine v2 (#3247)
• Add Pascal API for Moonshine v2 (#3248)
• Build huggingface spaces for Moonshine v2 with WebAssembly (#3249)

• Add Rust API for VAD (#3213)
• Replace deprecated std::istrstream with std::istringstream (#3214)
• Replace deprecated std::wstring_convert with manual UTF-8 codec (#3215)
• Fix CMake warnings: optional feature message level + policy version minimum (#3217)
• Upload FireRedASR2 CTC model (#3220)
• Bump hclust-cpp to 2026-02-25 release and modernize FetchContent (#3216)
• Support FireRedASR CTC models (#3221)
• Update language bindings for FireRedASR CTC models (#3224)

• Fix CI (#3192)
• Fix heap-buffer-overflow in ReadWaveImpl when data chunk size is odd (#3195)
• [PocketTTS] Add seed support and voice embedding caching for consiste… (#3189)
• Feat/pocket tts cache config (#3200)
• 3197: enhanced java binding for voice_embedding_cache_capacity (#3201)
• Dart, flutter, go, c-api binding and example (#3202)
• Begin to add Rust API (#3203)
• Add Rust API for streaming speech recognition (#3204)
• Add a real-time speech recognition example with microphone for Rust API. (#3205)
• Add Rust API for offline ASR (#3207)
• feat: Add PocketTTS cache & seed support to Node.js Addon and WASM APIs (#3206)
• Add more examples for offline ASR models with Rust API. (#3209)
• Update C#/Swift/Pascal API for PocketTTS' VoiceEmbeddingCacheCapacity. (#3211)

• Fix building without tts (#3168)
• Fix publishing npm packages for Linux aarch64 and wheels for macOS (#3169)
• Export PocketTTS for earlier versions of onnxruntime (#3170)
• Fix building wheels for Python 3.14 (#3182)
• Update Eigen from 3.4.0 to 3.4.1 (#3178)
• fix(flutter): add missing FFI struct fields for OfflineWhisper and FunAsrNano (#3186)
• Fix building wheels for Windows (#3187)

• Fix UnicodeDecodeError when accessing tokens in FunASR-nano tokenizer (#3058)
• Use more jobs for building VAD ASR APKs (#3068)
• Add export CGO_ENABLED=1 to all GO examples. (#3069)
• Support BPE tokenizer (#3078)
• Add C++ runtime and Python support PocketTTS for streaming voice cloning on CPU (#3083)
• Refactor addon loading logic and add static import for platform-specific binaries (#3075)
• Update C++ binary for PocketTTS (#3087)
• Add Python API examples for PocketTTS (#3088)
• Limit text length for PocketTTS. (#3089)
• Add CI for PocketTTS. (#3090)
• Fix Python CI (#3091)
• Fix build error (#3096)
• Add Java and Kotlin API for PocketTTS (#3095)
• Refactor JNI to remove casting. (#3103)
• Refactor JNI (#3107)
• Support MD and MT MSVC runtime libraries (CRT) for Windows x64 static build (#3111)
• Fix MSVC CRT for Windows x64 shared build. (#3114)
• Fix MSVC CRT for Windows arm64 (#3117)
• Fix MSVC CRT for Windows x86 (#3118)
• Refactor CI for Windows x64 (#3119)
• Fix CI for Windows x64 (#3123)
• Upload WenetSpeech-Wu u2pp ASR models. (#3125)
• Add TTS generation with GenerationConfig params C API (#3115)
• Refactor TTS C API (#3127)
• Add CXX API for PocketTTS (#3128)
• Add Swift API for PocketTTS (#3129)
• fix(android): Optimize UI updates and remove dead code in MainActivity (#3130)
• Change RPATH for sherpa-onnx.node (#3131)
• Add async js API for tts generate. (#3133)
• fix(android): Initialize models in background coroutine to avoid UI blocking (#3132)
• Add hotword support for FunASR-Nano (#3122)
• Provide async JS API to create TTS. (#3134)
• feat: Add a WebAssembly Text-to-Speech (TTS) demo with UI and worker-based audio generation using sherpa-onnx. (#3120)
• feat: add support for Meta Omnilingual ASR v2 models (#3138)
• Export omnilingualASR v2 (#3140)
• feat: Add ys_log_probs to NeMo transducer greedy search decoder (#3105)
• Add modified beam search and hotwords support for NeMo transducer models (#3077)
• Fix ORT Value default construction for Android build (#3141)
• Whisper timestamps (#2945)
• Add node-addon JavaScript API for PocketTTS (#3139)
• Update lifecycle-runtime-ktx version to 2.5.1 (#3143)
• Enable return value in callback for TTS in Go API. (#3150)
• Refactor Go API for TTS (#3151)
• Export models for CANN 8.1 (#3152)
• Add Go API for PocketTTS (#3153)
• Export models for CANN 8.3 and 8.5 (#3156)
• Add https://huggingface.co/alphacep/vosk-model-small-streaming-bn (#3158)
• Add Pascal API for Pocket TTS (#3157)
• Upload Vietnamese ASR models (#3159)
• Refactor Pascal API (#3160)
• Add C# API for PocketTTS. (#3162)
• Add JavaScript (WebAssembly) API for PocketTTS (#3163)
• Add Dart API for PocketTTS (#3164)
• Add GeneratedAudio ToBuffer() to the GO API (#3136)
• fix: resolve high vulnerability python.lang.security.audit.dangerous-system-call-tainted-env-args.dangerous-system-call-tainted-env-args (#3155)
• Fix various language bindings (#3166)

• Node addon api jsdoc (#3005)
• Add JavaScript async api for OfflineRecognizer decodeStream. (#3049)
• Support creating OfflineRecognizer asynchronously in JavaScript. (#3050)
• Fix uploading files to huggingface (#3054)
• Add Dart API for FunASR Nano (#3055)
• Fix uploading APK files (#3056)

• Update wav files for FunASR Nano (#3038)
• cmake: fix sha256 for onnxruntime linux x86_64 gpu package (#3042)
• Fix checking funasr nano tokenizer on Windows (#3043)
• Support nemotron-speech-streaming-en-0.6b (#3044)
• Build APK for nemotron-speech-streaming-en-0.6b (#3045)
• Fix building Linux arm wheels (#3047)

• Fix publishing NPM packages (#2909)
• Refactor ZipVoice C++ code (#2911)
• Export more zipformer ctc models to qnn (#2921)
• [KWS] Add phone+ppinyin tokenization with lexicon support (for zh-en model) (#2922)
• Export Paraformer ASR models to QNN (#2925)
• Add Transpose for a 2-D matrix. (#2926)
• Optimize computation with Eigen. (#2928)
• Add C++ runtime for Paraformer ASR models with Qualcomm NPU using QNN (#2931)
• Add Android demo for Paraformer ASR with Qualcomm NPU. (#2932)
• Export Google MedASR to sherpa-onnx (#2934)
• Add C++ runtime and Python API for Google MedASR models (#2935)
• Fix creating a view of an Ort::Value tensor. (#2939)
• Add C and CXX API for Google MedASR model (#2946)
• [TTS Engine] Fix engine speed (#2895)
• Add Swift API for Google MedASR model (#2947)
• Add C# API for Google MedASR model (#2949)
• Add Pascal API for Google MedASR model (#2950)
• Add Go API for Google MedAsr model (#2952)
• Add Dart API for Google MedAsr model (#2953)
• Add JavaScript API (WebAssembly) for Google MedAsr model (#2954)
• Add JavaScript API (node-addon) for Google MedAsr model (#2955)
• Add Kotlin and Java API for Google MedAsr model (#2956)
• Add funASR-Nano with LLM support (#2936)
• Fix building for Windows (#2964)
• Fix building for HarmonyOS (#2972)
• [feature] add FunASRNano config into golang api (#2974)
• Update FunAsr-Nano CTC model (#2978)
• [opt] opt free pointer function in Go API (#2975)
• [feature] use jinja2 to generate sherpa-onnx-go lib (#2976)
• Reformat Go API code (#2979)
• Fix building for onnxruntime >= 1.11.0 (#2981)
• Export Whisper to RK NPU (#2983)
• Test Whisper on Ascend NPU using ACL Python API (#2986)
• FunASR-nano: switch to unified KV-cache LLM (#2995)
• Remove filesystem header (#2998)
• Fix(csrc/melotts): Fix V-words pronunciation on MeloTTS_en (#3002)
• Upload FunASR Nano ASR models with LLM (#3003)
• Fix download test wav files (#3004)
• Use onnxruntime 1.23.2 for Windows (#3007)
• Add CI to export Whisper models to Ascend NPU (#3008)
• Add C++ runtime for Whisper with Ascend NPU (#3009)
• Use onnxruntime v1.23.2 for Linux aarch64 (#3016)
• Use onnxruntime v1.23.2 for Linux arm (#3017)
• Start to switch from onnxruntime 1.17.1 to v1.23.2 (#2993)
• Use onnxruntime 1.23.2 for Linux x64 + NVIDIA GPU (#3018)
• Update CI test for FunASR Nano C/C++ API (#3021)
• [feature] add FunASRNano Swift api (#2994)
• swift: add FunASR nano Swift API (#3022)
• Add Go API test for FunASR Nano (#3025)
• Add JavaScript API for FunASR Nano (node-addon) (#3026)
• Add Pascal API for FunASR Nano (#3029)
• Add C# API for FunASR Nano (#3031)
• Add Kotlin and Java API for FunASR Nano models (#3030)
• Fire-Red-ASR: enable ORT I/O binding for encoder/decoder (#3011)
• whisper: improve ORT IO binding execution (#3023)
• Add JavaScript API for FunASR Nano (WebAssembly) (#3027)
• Fix CI test for nodejs (#3033)

• Refactor axcl examples. (#2867)
• Update README to include Axera NPU (#2870)
• Add CI for Axera NPU (#2872)
• Refactor sense voice impl (#2873)
• Refactor Paraformer Impl (#2874)
• Remove unused lock file (#2875)
• Load QNN context binary for faster startup (#2877)
• Export models to Ascend 910B4 (#2878)
• Optimize streaming output results when VAD does not detect human voice for a long time (#2876)
• Build APKs for MatchaTTS Chinese+English (#2882)
• Publish WASM spaces for MatchaTTS Chinese+English model (#2885)
• Add script for testing zipvoice onnx models (#2887)
• upload zipvoice onnx models (#2890)
• Remove cppinyin from zipvoice (#2892)
• Fix building errors (#2893)
• Use a shorter name for Zipvoice models. (#2894)
• Export GigaAM v3 to sherpa-onnx (#2901)
• Fix typos in URL (#2905)
• Support Fun-ASR-Nano-2512 (#2906)

• Fix building without TTS for C API (#2838)
• [ZipVoice] Fix english tokenization error (#2834)
• Add simulate streaming ASR Python example for Paraformer (#2839)
• Fix building JNI for Windows (#2840)
• Avoid NaN in NeMo speaker embedding models. (#2844)
• Add spacemit ort ep for spacemit riscv cpus (#2837)
• Add token-level confidence scores (ys_probs) for offline transducer models (#2843)
• Fix token log probabilities in offline transducer modified beam search decoder (#2846)
• Support AXERA ax630, ax650, and axcl backends. (#2849)
• Refactor axera npu examples (#2850)
• Fix matcha tts zh-en model (#2851)
• Fix the English part for Matcha TTS. (#2853)
• Refactor text-utils (#2855)
• Fix matcha tts (#2856)
• Add a space between English words for Matcha zh-en TTS (#2858)
• Fix punctuations in matcha zh-en tts (#2859)
• Upload matcha tts zh-en model (#2865)
• Fix the discrepancy with the Silero VAD isSpeech logic (#2863)

• Fix building wheels (#2786)
• export omniASR_CTC_1B (#2788)
• Add C++ QNN support for SenseVoice (#2793)
• Export models for CANN toolkit 7.0 (#2795)
• Support hotwords with byte level bpe (#2802)
• Add Android demo with QNN (Qualcomm NPU) for SenseVoice ASR (#2803)
• Export zipformer ctc models to QNN (#2815)
• Add spaces between English words for Homophone replacer. (#2817)
• Add C++ QNN support for Zipformer CTC models. (#2809)
• Limit symbol visibility in the shared libraries (#2822)
• Fix warnings for initializing tts lexicon. (#2823)
• Export zipformer ctc models to Ascend NPU (#2824)
• Refactor scripts for exporting models to Ascend NPU. (#2825)
• Add C++ support for Zipformer CTC on Ascend NPU (#2826)
• Fix segfault when non-wav file is passed to ReadWave (#2821)
• Avoid calling rknn_dup_context(). (#2828)
• Add C++ support for Paraformer with RK NPU (#2829)
• Update README to include NPU support (#2830)
• Support running whisper large v3 with external data weight (#2807)

• Fix releasing

• Exposing online punctuation model support in node-addon-api (#2609)
• Fix building wheels (#2619)
• Export one more Piper Arabic TTS model (#2623)
• fix: hot update language for sencevoice (#2627)
• Add C API and Go API for Zipvoice (#2628)
• Add CI tests for Zipvoice Go API (#2630)
• Remove hardcoded dithering value in NeMo transducer recognizer (#2639)
• Reduce verbose output about reading lexicon for TTS (#2648)
• Add Parakeet TDT model for generating subtitles (#2649)
• Add more Piper TTS models (#2651)
• Add CXX API for audio tagging (#2652)
• Add C# API for audio tagging (#2653)
• Support KWS + RKNN. (#2190)
• Support https://github.com/ASLP-lab/WenetSpeech-Chuan (#2656)
• Fix building for android (#2657)
• fix ios build script (#2645)
• Update kaldi-native-fbank (#2659)
• Add missing python class definitions for builds without TTS support (#2660)
• Remove jieba from kokoro and matcha tts. (#2662)
• add flet_sherpa_onnx in readme (#2663)
• Remove cppjieba (#2664)
• Add phrase matcher to merge words into phrases for TTS. (#2668)
• Limit number of tokens per sentence in MatchaTTS. (#2671)
• Update README to include a ROS2 project using sherpa-onnx (#2672)
• Fix building Flutter APPs (#2673)
• Export Paraformer to RKNN (#2689)
• Update README.md add achatbot-go Projects using sherpa-onnx link (#2691)
• Add CI to export Paraformer to RKNN (#2692)
• Support MatchaTTS with English and Chinese (#2695)
• Export Paraformer ASR models from FunASR to Ascend NPU 910B (#2697)
• Update README to include Ascend NPU (#2698)
• Fix WASM (JS) after adding zipvoice. (#2702)
• Export SenseVoice ASR models to Ascend NPU 910B (#2707)
• Fix building for various language bindings after adding zipvoice (#2709)

• Fix setting rknn core mask (#2594)
• Add Dart API for spoken language identification (#2596)
• Add CI tests for dart spoken language identification example (#2598)
• Provide pre-compiled sherpa-onnx libs/binaries for CUDA 12.x + onnxruntime 1.22.0 (#2599)
• Provide pre-compiled whls for cuda 12.x on Linux x64 and Windows x64 (#2601)
• Fix TDT decoding for NeMo TDT transducers (#2606)
• Add a C++ example for simulated streaming ASR (#2607)

• Fix initializing symbol table for OnlineRecognizer. (#2590)
• Support RK NPU for SenseVoice non-streaming ASR models (#2589)
• Upload RKNN models for sense-voice (#2592)

• Fix building for risc-v (#2549)
• Fix using sherpa-onnx as a cmake sub-project. (#2550)
• Update kaldifst and kaldi-decoder (#2551)
• Support armv8l in Java API (#2556)
• Disable loading libs from jar on Android. (#2557)
• Fix cantonese vits tts (#2558)
• Avoid appending blanks for Cantonese vits tts. (#2559)
• Add hint for loading model files from SD card on Android. (#2564)
• Update README to include https://github.com/Mentra-Community/MentraOS (#2565)
• Export models from https://github.com/voicekit-team/T-one to sherpa-onnx (#2571)
• Add C++ and Python support for T-one streaming Russian ASR models (#2575)
• Add various language bindings for streaming T-one Russian ASR models (#2576)
• Fix the missing online punctuation in android aar (#2577)
• Export KittenTTS mini v0.1 to sherpa-onnx (#2578)
• Upload new sense-voice models (#2580)
• Export ASLP-lab/WSYue-ASR/tree/main/u2pp_conformer_yue to sherpa-onnx (#2582)
• Add various language bindings for Wenet non-streaming CTC models (#2584)

• Add two more Piper tts models (#2525)
• Generate tts samples for MatchaTTS (English). (#2527)
• Fix releasing go packages (#2529)
• Add license info about tts models from OpenVoiceOS (#2530)
• Support BPE models with byte fallback. (#2531)
• Simplify the usage of our non-Android Java API (#2533)
• Fix wasm for kws (#2535)
• Add one more German tts model from OpenVoiceOS. (#2536)
• Fix uploading win32 libs to huggingface (#2537)
• Add Zipvoice (#2487)
• Fix c api (#2545)
• Fix linking (#2546)

• Add VOSK streaming Russian ASR models and Kroko streaming German ASR models (#2502)
• Refactor CI tests (#2504)
• Update APK versions (#2505)
• Export whisper distil-large-v3 and distil-large-v3.5 to sherpa-onnx (#2506)
• Support specifying pronunciations of phrases in Chinese TTS. (#2507)
• fix(flutter): fix unicode problem in windows path (#2508)
• feat: add punctuation C++ API (#2510)
• Fix ctrl+c may lead to coredump (#2511)
• Add kitten tts nano v0.2 (#2512)
• Scripts to generate tts samples (#2513)
• Add tdt duration to APIs (#2514)
• Support 16KB page size for Android (#2520)
• Split sherpa-onnx Python package (#2521)
• Fix kokoro tts for punctuations (#2522)

• Add more piper tts models (#2480)
• Fix ASR for UE (#2483)
• push to maven center (#2463)
• Specify ABIs when building APKs (#2488)
• Add more debug info for vits tts (#2491)
• Add Swift API for computing speaker embeddings (#2492)
• Alex/feat add python example (#2490)
• Support TDT transducer decoding (#2495)
• Fix java test (#2496)
• Refactor Swift API (#2493)
• add TtsReader app to README.md (#2498)
• Export https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3 to sherpa-onnx (#2500)
• Fix building apk (#2499)

• Expose JNI to compute probability of chunk in VAD (#2433)
• Add https://huggingface.co/Banafo/Kroko-ASR (#2453)
• Add APIs for Online NeMo CTC models (#2454)
• Export https://github.com/KittenML/KittenTTS to sherpa-onnx (#2456)
• Fix punctuations in kokoro tts. (#2458)
• Limit number of tokens in fire red asr decoding. (#2459)
• Add C++ runtime for kitten-tts (#2460)
• Add Kotlin and Java API for KittenTTS (#2461)
• Add Android TTS Engine APK for KittenTTS (#2465)
• Add Python API for KittenTTS. (#2466)
• Add C API for KittenTTS (#2467)
• Add CXX API for KittenTTS (#2469)
• Add JavaScript API (node-addon) for KittenTTS (#2470)
• Add JavaScript API (WebAssembly) for KittenTTS (#2471)
• Add Pascal API for KittenTTS (#2474)
• Add Dart API for KittenTTS (#2475)
• Add Swift API for KittenTTS (#2476)
• Add C# API for KittenTTS (#2477)
• Add Go API for KittenTTS (#2478)

• Support Portuguese and German ASR models from NeMo (#2394)
• Support returning the current speech segment for VAD. (#2397)
• Add more piper tts polish models (#2403)
• Support VAD+ASR for WearOS (#2404)
• Support test long audio with streaming-model & vad (#2405)
• Fix typo in sherpa-onnx-vad-with-online-asr.cc (#2407)
• Add tail padding for sherpa-onnx-vad-with-online-asr (#2408)
• Add more French TTS models (#2424)
• Add more piper tts models (#2425)
• Implement max_symbols_per_frame for GigaAM2 accurate decoding since model uses char tokens instead of BPE. (#2423)
• Fix GigaAM transducer encoder output length data type (#2426)
• Add friendly log messages for Android and HarmonyOS TTS users. (#2427)
• Fix setGraph in OnlineCtcFstDecoderConfig Java API (#2411)

• Support silero-vad v4 exported by k2-fsa (#2372)
• Add C++ and Python support for ten-vad (#2377)
• Fix compile errors for Linux (#2378)
• Add C API for ten-vad (#2379)
• Add CXX API examples for ten-vad. (#2380)
• Add JavaScript (WebAssembly) API for ten-vad (#2382)
• Add JavaScript (node-addon) API for ten-vad (#2383)
• Add Go API for ten-vad (#2384)
• Add C# API for ten-vad (#2385)
• Add Dart API for ten-vad (#2386)
• Add Swift API for ten-vad (#2387)
• Add Pascal API for ten-vad (#2388)
• Add Java/Kotlin API and Android support for ten-vad (#2389)

• Fix typo CMAKE_EXECUTBLE_LINKER_FLAGS -> CMAKE_EXECUTABLE_LINKER_FLAGS (#2344)
• Fix testing dart packages (#2345)
• fix(canary): use dynamo export, single input_ids and avoid 0/1 specialization (#2348)
• Fix TTS for Unreal Engine (#2349)
• Update readme to include https://github.com/mawwalker/stt-server (#2350)
• Add meta data to NeMo canary ONNX models (#2351)
• Update README to include https://github.com/bbeyondllove/asr_server (#2353)
• Add C++ runtime and Python API for NeMo Canary models (#2352)
• Add C/CXX/JavaScript API for NeMo Canary models (#2357)
• Add Java and Kotlin API for NeMo Canary models (#2359)
• Upload fp16 onnx model files for FireRedASR (#2360)
• Fix nemo feature normalization in test code (#2361)
• Refactor exporting NeMo models (#2362)
• Add LODR support to online and offline recognizers (#2026)
• Add CXX examples for NeMo TDT ASR. (#2363)
• Add Pascal/Go/C#/Dart API for NeMo Canary ASR models (#2367)

• Refactor release scripts. (#2323)
• Add TTS engine APKs for more models (#2327)
• Fix static link without tts (#2328)
• Fix VAD+ASR C++ example. (#2335)
• Add sherpa-onnx-streaming-zipformer-zh-int8-2025-06-30 to android ASR apk (#2336)
• Support non-streaming zipformer CTC ASR models (#2340)
• Support linux aarch64 for Dart and Flutter (#2342)

• Show CMake debug information. (#2316)
• Remove portaudio-go in Go API examples. (#2317)
• Support Zipformer CTC ASR with whisper features. (#2319)
• Support Zipformer transducer ASR with whisper features. (#2321)

• Fix CI for windows (#2279)
• Add jar for Java 24. (#2280)
• Add Python API for source separation (#2283)
• Add link to huggingface space for source separation. (#2284)
• Fix isspace on windows in debug build (#2042)
• Update wasm/vad-asr/assets/README.md for more clear (#2297)
• Update TTS Engine APK to support multi-lang (#2294)
• Add scripts for exporting Piper TTS models to sherpa-onnx (#2299)
• Update sherpa-onnx-shared.pc.in (#2300)
• Fixes #2172 (#2301)
• Refactor kokoro export (#2302)
• Fix building for Pascal (#2305)
• Support extra languages in multi-lang kokoro tts (#2303)
• Update readme to include BreezeApp from MediaTek Research. (#2313)
• Add API to get version information (#2309)

• Use jlong explicitly in jni. (#2229)
• Fix building RKNN wheels (#2233)
• Fix publishing binaries for RKNN (#2234)
• Export spleeter model to onnx for source separation (#2237)
• Add C++ runtime for spleeter about source separation (#2242)
• Add include headers for ANDROID_API,OHOS (#2251)
• JAVA-API: Manual Library Loading Support for Restricted Environments (#2253)
• Build APK with replace.fst (#2254)
• repair rknn wheels (#2257)
• Update kaldi-native-fbank. (#2259)
• Fix building sherpa-onnx (#2262)
• Fix building MFC examples (#2263)
• Add UVR models for source separation. (#2266)
• move portaudio common record code to microphone (#2264)
• fixed mfc build error (#2267)
• Add C++ support for UVR models (#2269)
• Export nvidia/canary-180m-flash to sherpa-onnx (#2272)
• Update utils.dart (#2275)
• Fix rknn for multi-threads (#2274)
• Fix 32-bit arm CI (#2276)

• Fix building wheels for macOS (#2192)
• Show verbose logs in homophone replacer (#2194)
• Fix displaying streaming speech recognition results for Python. (#2196)
• Add real-time speech recognition example for SenseVoice. (#2197)
• docs: add Open-XiaoAI KWS project (#2198)
• Add C++ example for streaming ASR with SenseVoice. (#2199)
• Add C++ example for real-time ASR with nvidia/parakeet-tdt-0.6b-v2. (#2201)
• Add a link to YouTube video including sherpa-onnx. (#2202)
• Support sending is_eof for online websocket server. (#2204)
• Add alsa-based streaming ASR example for sense voice. (#2207)
• Support homophone replacer in Android asr demo. (#2210)
• Add Go implementation of the TTS generation callback (#2213)
• Add Android demo for real-time ASR with non-streaming ASR models. (#2214)
• Expose dither for JNI (#2215)
• Add nodejs example for parakeet-tdt-0.6b-v2. (#2219)
• Add script to build APK for simulated-streaming-asr. (#2220)

• export parakeet-tdt-0.6b-v2 to sherpa-onnx (#2180)
• Add C++ runtime for parakeet-tdt-0.6b-v2. (#2181)
• Avoid NaN in feature normalization. (#2186)

• Disable strict hotword matching mode for offline transducer (#1837)
• Comment refinement: Add note about vocoder file for matcha TTS config (#2106)
• Fix a typo in the JNI for Android. (#2108)
• Generate subtitles with FireRedAsr models (#2112)
• Use manylinux_2_28_x86_64 to build linux gpu for sherpa-onnx (#2123)
• Support running sherpa-onnx with RK NPU on Android (#2124)
• Fix building for HarmonyOS (#2125)
• cmake build, configurable from env (#2115)
• Expose dither in python API (#2127)
• Add support for GigaAM-CTC-v2 (#2135)
• Support Giga AM transducer V2 (#2136)
• Export kokoro 1.0 int8 models (#2137)
• Upload more onnx ASR models (#2141)
• Fix building for open harmonyOS (#2142)
• online-transducer: reset the encoder together with 2 previous output symbols (non-blank) (#2129)
• Fix punctuations for kokoro tts 1.1-zh. (#2146)
• Fix setting OnlineModelConfig in Java API (#2147)
• Support decoding multiple streams in Java API. (#2149)
• Support replacing homophonic phrases (#2153)
• Add C and CXX API for homophone replacer (#2156)
• Add JavaScript API (WASM) for homophone replacer (#2157)
• Add JavaScript API (node-addon) for homophone replacer (#2158)
• Fix building without TTS (#2159)
• Add homophone replacer example for Python API. (#2161)
• More fix for building without tts (#2162)
• Add Swift API for homophone replacer. (#2164)
• Add C# API for homophone replacer (#2165)
• Add Kotlin and Java API for homophone replacer (#2166)
• Add Dart API for homophone replacer (#2167)
• Add Go API for homophone replacer (#2168)

• fix vits dict dir config (#2036)
• fix case (#2037)
• Fix building wheels for RKNN (#2041)
• Change scale factor to 32767 (#2056)
• Fix length scale for kokoro tts (#2060)
• Allow building repository as CMake subdirectory (#2059)
• Export silero_vad v4 to RKNN (#2067)
• fix dml with preinstall ort (#2066)
• Fix building aar to include speech denoiser (#2069)
• Add CXX API for VAD (#2077)
• Add C++ runtime for silero_vad with RKNN (#2078)
• Refactor rknn code (#2079)
• Fix building for android (#2081)
• Add C++ and Python API for Dolphin CTC models (#2085)
• Add Kotlin and Java API for Dolphin CTC models (#2086)
• Add C and CXX API for Dolphin CTC models (#2088)
• Preserve more context after endpointing in transducer (#2061)
• Add C# API for Dolphin CTC models (#2089)
• Add Go API for Dolphin CTC models (#2090)
• Add Swift API for Dolphin CTC models (#2091)
• Add Javascript (WebAssembly) API for Dolphin CTC models (#2093)
• Add Javascript (node-addon) API for Dolphin CTC models (#2094)
• Add Dart API for Dolphin CTC models (#2095)
• Add Pascal API for Dolphin CTC models (#2096)

• Fix CI (#2016)
• Publish jar for more java versions (#2017)
• add alsa example for vad+offline asr (#2020)
• Support cuda12 and cudnn8 for Linux aarch64. (#2021)
• Update README to include more projects using sherpa-onnx (#2022)
• Fix a bug in vad.reset() (#2023)
• Fix Matcha + vocos for Android (#2024)
• Fix crash in Android tts engine demo. (#2029)
• Fix build script: add 'cd build' after 'mkdir build' to ensure the correct working directory for CMake (#2033)
• fix static linking (#2032)

• Export vocos to sherpa-onnx (#2012)
• Add C++ runtime for vocos (#2014)

• Fix building wheels for Python 3.7 (#1933)
• Add Kotlin and Java API for online punctuation models (#1936)
• Add Kokoro v1.1-zh (#1942)
• Support RKNN for Zipformer CTC models. (#1948)
• Add transducer modified_beam_search for RKNN. (#1949)
• Update README to include projects that is using sherpa-onnx (#1956)
• Limit number of tokens per second for whisper. (#1958)
• Ebranchformer (#1951)
• Test using sherpa-onnx as a cmake subproject (#1961)
• Add C++ demo for VAD+non-streaming ASR (#1964)
• Export gtcrn models to sherpa-onnx (#1975)
• c-api add wave write to buffer. (#1962)
• add SherpaOnnxOfflineRecognizerSetConfig binding for go, and optimize the new/free for C.struct_SherpaOnnxOfflineRecognizerConfig ptr (#1976)
• Add C++ runtime for speech enhancement GTCRN models (#1977)
• Add Python API for speech enhancement GTCRN models (#1978)
• Add C API for speech enhancement GTCRN models (#1984)
• Add CXX API for speech enhancement GTCRN models (#1986)
• Add Swift API for speech enhancement GTCRN models (#1989)
• Add C# API for speech enhancement GTCRN models (#1990)
• Add Go API for speech enhancement GTCRN models (#1991)
• Add Pascal API for speech enhancement GTCRN models (#1992)
• Add Dart API for speech enhancement GTCRN models (#1993)
• Add JavaScript (node-addon) API for speech enhancement GTCRN models (#1996)
• Add WebAssembly (WASM) for speech enhancement GTCRN models (#2002)
• Add JavaScript API (wasm) for speech enhancement GTCRN models (#2007)
• Add Kotlin API for speech enhancement GTCRN models (#2008)
• Add Java API for speech enhancement GTCRN models (#2009)

• Fix kokoro lexicon. (#1886)
• speaker-identification-with-vad-non-streaming-asr.py Lack of support for sense_voice. (#1884)
• Fix generating Chinese lexicon for Kokoro TTS 1.0 (#1888)
• Reduce vad-whisper-c-api example code. (#1891)
• JNI Exception Handling (#1452)
• Fix #1901: UnicodeEncodeError running export_bpe_vocab.py (#1902)
• Fix publishing pre-built windows libraries (#1905)
• Fixing Whisper Model Token Normalization (#1904)
• feat: add mic example for better compatibility (#1909)
• Add onnxruntime 1.18.1 for Linux aarch64 GPU (#1914)
• Add C++ API for streaming zipformer ASR on RK NPU (#1908)
• change [1<<28] to [1<<10], to fix build issues on GOARCH=386 that [1<<28] too large (#1916)
• Flutter Config toJson/fromJson (#1893)
• Fix publishing linux pre-built artifacts (#1919)
• go.mod set to use go 1.17, and use unsafe.Slice to optimize the code (#1920)
• fix: AddPunct panic for Go(#1921)
• Fix publishing macos pre-built artifacts (#1922)
• Minor fixes for rknn (#1925)
• Build wheels for rknn linux aarch64 (#1928)

• [update] fixed bug: create golang instance succeed while the c struct create failed (#1860)
• fixed typo in RTF calculations (#1861)
• Export FireRedASR to sherpa-onnx. (#1865)
• Add C++ and Python API for FireRedASR AED models (#1867)
• Add Kotlin and Java API for FireRedAsr AED model (#1870)
• Add C API for FireRedAsr AED model. (#1871)
• Add CXX API for FireRedAsr (#1872)
• Add JavaScript API (node-addon) for FireRedAsr (#1873)
• Add JavaScript API (WebAssembly) for FireRedAsr model. (#1874)
• Add C# API for FireRedAsr Model (#1875)
• Add C# API for FireRedAsr Model (#1875)
• Add Swift API for FireRedAsr AED Model (#1876)
• Add Dart API for FireRedAsr AED Model (#1877)
• Add Go API for FireRedAsr AED Model (#1879)
• Add Pascal API for FireRedAsr AED Model (#1880)

• Export MatchaTTS fa-en model to sherpa-onnx (#1832)
• Add C++ support for MatchaTTS models not from icefall. (#1834)
• OfflineRecognizer supports create stream with hotwords (#1833)
• Add PengChengStarling models to sherpa-onnx (#1835)
• Support specifying voice in espeak-ng for kokoro tts models. (#1836)
• Fix: made print sherpa_onnx_loge when it is in debug mode (#1838)
• Add Go API for audio tagging (#1840)
• Fix CI (#1841)
• Update readme to contain links for pre-built Apps (#1853)
• Modify the model used (#1855)
• Flutter OnlinePunctuation (#1854)
• Fix spliting text by languages for kokoro tts. (#1849)

• Add MFC example for Kokoro TTS 1.0 (#1815)
• Update sherpa-onnx-tts.js VitsModelConfig.model can be none (#1817)
• Fix passing gb2312 encoded strings to tts on Windows (#1819)
• Support scaling the duration of a pause in TTS. (#1820)
• Fix building wheels for linux aarch64. (#1821)
• Fix CI for Linux aarch64. (#1822)

• Fix publishing wheels (#1746)
• Update README to include https://github.com/xinhecuican/QSmartAssistant (#1755)
• Add Kokoro TTS to MFC examples (#1760)
• Refactor node-addon C++ code. (#1768)
• Add keyword spotter C API for HarmonyOS (#1769)
• Add ArkTS API for Keyword spotting. (#1775)
• Add Flutter example for Kokoro TTS (#1776)
• Initialize the audio session for iOS ASR example (#1786)
• Fix: Prepend 0 to tokenization to prevent word skipping for Kokoro. (#1787)
• Export Kokoro 1.0 to sherpa-onnx (#1788)
• Add C++ and Python API for Kokoro 1.0 multilingual TTS model (#1795)
• Add Java and Kotlin API for Kokoro TTS 1.0 (#1798)
• Add Android demo for Kokoro TTS 1.0 (#1799)
• Add C API for Kokoro TTS 1.0 (#1801)
• Add CXX API for Kokoro TTS 1.0 (#1802)
• Add Swift API for Kokoro TTS 1.0 (#1803)
• Add Go API for Kokoro TTS 1.0 (#1804)
• Add C# API for Kokoro TTS 1.0 (#1805)
• Add Dart API for Kokoro TTS 1.0 (#1806)
• Add Pascal API for Kokoro TTS 1.0 (#1807)
• Add JavaScript API (node-addon) for Kokoro TTS 1.0 (#1808)
• Add JavaScript API (WebAssembly) for Kokoro TTS 1.0 (#1809)
• Add Flutter example for Kokoro TTS 1.0 (#1810)
• Add iOS demo for Kokoro TTS 1.0 (#1812)
• Add HarmonyOS demo for Kokoro TTS 1.0 (#1813)

• Fix UI for Android TTS Engine. (#1735)
• Add iOS TTS example for MatchaTTS (#1736)
• Add iOS example for Kokoro TTS (#1737)
• Fix dither binding in Pybind11 to ensure independence from high_freq in FeatureExtractorConfig (#1739)
• Fix keyword spotting. (#1689)
• Update readme to include https://github.com/hfyydd/sherpa-onnx-server (#1741)
• Reduce vad-moonshine-c-api example code. (#1742)
• Support Kokoro TTS for HarmonyOS. (#1743)

• Fix building wheels (#1703)
• Export kokoro to sherpa-onnx (#1713)
• Add C++ and Python API for Kokoro TTS models. (#1715)
• Add C API for Kokoro TTS models (#1717)
• Fix style issues (#1718)
• Add C# API for Kokoro TTS models (#1720)
• Add Swift API for Kokoro TTS models (#1721)
• Add Go API for Kokoro TTS models (#1722)
• Add Dart API for Kokoro TTS models (#1723)
• Add Pascal API for Kokoro TTS models (#1724)
• Add JavaScript API (node-addon) for Kokoro TTS models (#1725)
• Add JavaScript (WebAssembly) API for Kokoro TTS models. (#1726)
• Add Kotlin and Java API for Kokoro TTS models (#1728)
• Update README.md for KWS to not use git lfs. (#1729)

• Fix building without TTS (#1691)
• Add README for android libs. (#1693)
• Fix: export-onnx.py(expected all tensors to be on the same device) (#1699)
• Fix passing strings from C# to C. (#1701)

• Fix initializing TTS in Python. (#1664)
• Remove spaces after punctuations for TTS (#1666)
• Add constructor fromPtr() for all flutter class with factory ctor. (#1667)
• Add Kotlin API for Matcha-TTS models. (#1668)
• Support Matcha-TTS models using espeak-ng (#1672)
• Add Java API for Matcha-TTS models. (#1673)
• Avoid adding tail padding for VAD in generate-subtitles.py (#1674)
• Add C API for MatchaTTS models (#1675)
• Add CXX API for MatchaTTS models (#1676)
• Add JavaScript API (node-addon-api) for MatchaTTS models. (#1677)
• Add HarmonyOS examples for MatchaTTS. (#1678)
• Upgraded to .NET 8 and made code style a little more internally consistent. (#1680)
• Update workflows to use .NET 8.0 also. (#1681)
• Add C# and JavaScript (wasm) API for MatchaTTS models (#1682)
• Add Android demo for MatchaTTS models. (#1683)
• Add Swift API for MatchaTTS models. (#1684)
• Add Go API for MatchaTTS models (#1685)
• Add Pascal API for MatchaTTS models. (#1686)
• Add Dart API for MatchaTTS models (#1687)