fquirin / speech-recognition-experiments

Experiments to test different speech recognition systems for SEPIA Framework
MIT License
57 stars 5 forks source link

Request to add sherpa-onnx #3

Open csukuangfj opened 1 year ago

csukuangfj commented 1 year ago

Please see https://k2-fsa.github.io/sherpa/onnx/index.html for the documentation

It is similar to sherpa-ncnn but uses onnxruntime instead of ncnn for neural network computation.

fquirin commented 1 year ago

I've put it on my to-do list 👍

kmn1024 commented 9 months ago

The same team now has pre-builts for many Whisper sizes too! https://k2-fsa.github.io/sherpa/onnx/pretrained_models/whisper/export-onnx.html#available-models

Wonder how their performance compares.

csukuangfj commented 9 months ago

https://github.com/k2-fsa/sherpa-onnx/pull/471 The above PR removes the 30-second constraint from Whisper, which means you don't have to pad the input to 30-second.

It would be nice if someone could test it and compare the performance with other implementations.