Starstruk Logo MarkStarstruk

v1.10.29

k2-fsa/sherpa-onnxv1.10.29Oct 25, 2024by csukuangfj

AI Summary

Introduces Russian ASR support via GigaAM models and speaker diarization capabilities, alongside various API updates.

Key Highlights

  • Support for GigaAM CTC and NeMo transducer models for Russian ASR.
  • C++ APIs for streaming and non-streaming ASR.
  • Support for speaker diarization and speaker identification.
  • Go API for offline punctuation models.

New Features

  • GigaAM CTC models for Russian ASR
  • GigaAM NeMo transducer model for Russian ASR
  • C++ API for streaming ASR
  • C++ API for non-streaming ASR
  • Speaker diarization support
  • Speaker identification with VAD and non-streaming ASR

Full Release Notes

## What's Changed
* Upload speaker embedding models to huggingface by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1428
* "Speaker identification" is repeat! by @semxum in https://github.com/k2-fsa/sherpa-onnx/pull/1431
* Add Go API for offline punctuation models by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1434
* updated onnxruntime-linux-aarch64.cmake so that libonnxruntime.so can… by @shawl336 in https://github.com/k2-fsa/sherpa-onnx/pull/1436
* Support https://huggingface.co/Revai/reverb-diarization-v1 by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1437
* fix "log10" compile error by import CMATH lib by @Zazzle516 in https://github.com/k2-fsa/sherpa-onnx/pull/1438
* add more models for speaker diarization by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1440
* Add Java API example for hotwords. by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1442
* update java for hotword jar by @YeyuchenBa in https://github.com/k2-fsa/sherpa-onnx/pull/1444
* add java android demo by @JameWade in https://github.com/k2-fsa/sherpa-onnx/pull/1454
* Add C++ API for streaming ASR. by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1455
* Add C++ API for non-streaming ASR by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1456
* Fix style issues by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1458
* Handle NaN embeddings in speaker diarization. by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1461
* Add speaker identification with VAD and non-streaming ASR using ALSA by @Peakyxh in https://github.com/k2-fsa/sherpa-onnx/pull/1463
* Support GigaAM CTC models for Russian ASR by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1464
* Add GigaAM NeMo transducer model for Russian ASR by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1467
* Release v1.10.29 by @csukuangfj in https://github.com/k2-fsa/sherpa-onnx/pull/1468

## New Contributors
* @semxum made their first contribution in https://github.com/k2-fsa/sherpa-onnx/pull/1431
* @Zazzle516 made their first contribution in https://github.com/k2-fsa/sherpa-onnx/pull/1438
* @YeyuchenBa made their first contribution in https://github.com/k2-fsa/sherpa-onnx/pull/1444
* @JameWade made their first contribution in https://github.com/k2-fsa/sherpa-onnx/pull/1454
* @Peakyxh made their first contribution in https://github.com/k2-fsa/sherpa-onnx/pull/1463

**Full Changelog**: https://github.com/k2-fsa/sherpa-onnx/compare/v1.10.28...v1.10.29