RealtimeSTT icon indicating copy to clipboard operation
RealtimeSTT copied to clipboard

Published 20 hours ago • verified publisher icon KoljaB

Handling of pre-recorded audio and float16 to float32 conversion on CPU

Open amalvinob007 opened this issue 7 months ago • 1 comments

I am using the RealtimeSTT library to process pre-recorded PCM audio chunks. However, when feeding the audio, the system seems to expect live audio input and outputs the message "speak now", instead of processing the provided audio file. Additionally, I see a warning that the model's weights are automatically converted from float16 to float32 due to CPU usage, which might impact performance.

"[2025-03-31 10:21:51.636] [ctranslate2] [thread 2156] [warning] The compute type inferred from the saved model is float16, but the target device or backend do not support efficient float16 computation. The model weights have been automatically converted to use the float32 compute type instead. / speak now"

amalvinob007 avatar Mar 31 '25 04:03 amalvinob007