Real-time Inference Parameter Adjustment (Indexing Rate/Silence Protection/Segment Size) via the Command-Line Interface

[ke1298v]

I've noticed that when running RVC in real-time using Python scripts (not the GUI), the audio quality initially seems unstable. However, after adjusting certain parameters during runtime (such as indexing rate, silent protection, and segment size), the output suddenly becomes very stable and natural.

I've seen real-time logs such as:

• Finite Element Analysis Time
• Baseline Frequency Time
• Indexing Time
• Model Time

Is there a formal or recommended way to dynamically adjust these parameters during real-time inference (at the command-line/script level), or are they supposed to be fixed before runtime?

Also:

Does the GUI limit some real-time adjustment possibilities compared to using Python directly?