Stream Speed in Ubuntu

[RoboEvangelist]

The stable speed for stream is about --steps 900ms. Loops faster than that won't allow whisper.cpp to recognize what i say. Is there a way to accelerate the processing speed on Ubuntu (e.g., --steps faster than 900ms)? In MacOS I can go down to --steps 500ms, and whisper can still recognize my speech

[ggerganov]

Here a few things to try:

• use smaller model
• use quantized model
• use greedy decoding
• use GPU backend