INNER CODE UNIT · Python

rtf

voicepaw/so-vits-svc-fork · src/so_vits_svc_fork/inference/main.py:232

        rtf = t.elapsed / block_seconds
        LOG.info(f"Realtime inference time: {t.elapsed:.3f}s, RTF: {rtf:.3f}")
        if rtf > 1:
            LOG.warning("RTF is too high, consider increasing block_seconds")

    try:
        with sd.Stream(
            device=(input_device, output_device),
            channels=1,
            callback=callback,
            samplerate=svc_model.target_sample,
            blocksize=int(block_seconds * svc_model.target_sample),
            latency="low",
        ) as stream:
            LOG.info(f"Latency: {stream.latency}")
            while True:
                sd.sleep(1000)
    finally:

View source record →