Qualcomm NPU (QNN): APKs for streaming speech recognition
This page lists the streaming speech recognition APKs for sherpa-onnx,
one of the deployment frameworks of the Next-gen Kaldi project.
It uses Qualcomm NPU for computation, with HTP backend. You need a Phone with Qualcomm SM8350 chips or more powerful chips, e.g.
SM8450, SM8475, SM8550, SM8650, SM8750, SM8850, etc.
The name of an APK has the following rule:
- sherpa-onnx-{version}-qnn-{arch}-asr-{lang}-{model}.apk
where
- version: It specifies the current version, e.g., 1.13.2
- arch: The architecture targeted by this APK, e.g., arm64-v8a, armeabi-v7a, x86_64, x86
- qnn: It means qnn is used
- lang: The lang of the model used in the APK, e.g., en for English, zh for Chinese
- model: The name of the model used in the APK
You can download all supported models from
https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models-qnn
If your are interested in the source code for the APK, please visit
https://github.com/k2-fsa/sherpa-onnx/tree/master/android/SherpaOnnx
Note about the license The code of Next-gen Kaldi is using
Apache-2.0 license. However,
we support models from different frameworks. Please check the license of your selected model.
For Chinese users, please visit this address,
which replaces huggingface.co with hf-mirror.com
中国用户, 请访问这个地址
sherpa-onnx-1.13.2-qnn-arm64-v8a-asr-zh_en-x_asr_streaming_zipformer_transducer_punct_2026_06_05_chunk_480ms.apk
sherpa-onnx-1.13.2-qnn-arm64-v8a-asr-zh_en-x_asr_streaming_zipformer_transducer_2026_06_05_chunk_480ms.apk
sherpa-onnx-1.13.2-qnn-arm64-v8a-asr-zh_en-streaming_zipformer_transducer_2023_03_20_chunk_32.apk
sherpa-onnx-1.13.2-qnn-arm64-v8a-asr-zh_en-SM8850_x_asr_streaming_zipformer_transducer_punct_2026_06_05_chunk_480ms.apk
sherpa-onnx-1.13.2-qnn-arm64-v8a-asr-zh_en-SM8850_x_asr_streaming_zipformer_transducer_2026_06_05_chunk_480ms.apk