Intel

/

whisper-large-v2-onnx-int4-inc

like 23

Hi @Salama1429 , this is the demo for audio sample inference. For real-time transcription, I think you can record and run model at a certain frequency.

@MengniWang can you tell inference script for this ? cant't seem to find app.py

Hi @haiderasad , is https://github.com/intel/intel-extension-for-transformers/tree/main/examples/huggingface/onnxruntime/speech-recognition/quantization#3-audio-inference helpful?

thanks @MengniWang can you shed some light on benchmarking with WHisperX and Faster Whisper compared to this int8 implementation on the large-v2 model? basically looking for ways to scale whisper for huge audios inflow