实时语音转换出来的都是断断续续的哑音

[360PB]

使用的默认配置:python real-time-gui.py

Using fp16: True
Warning: Skipped loading some keys due to shape mismatch: {'estimator.input_pos'}
cfm loaded
length_regulator loaded
funasr version: 1.2.3.
Check update of funasr, and it would cost few times. You may disable it by set disable_update=True in AutoModel
You are using the latest version of funasr-1.2.3
2025-02-10 20:53:31,567 - modelscope - INFO - Use user-specified model revision: v2.0.4
Input device: 1:麦克风 (Realtek(R) Audio)
Output device: 20:Voicemeeter Input (VB-Audio Voi
cuda_is_available: True
(3969, 2)
rtf_avg: 0.293: 100%|████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 23.44it/s]
[]
Time taken for VAD: 43.56163024902344ms
preprocess time: 0.09
Setting ce_dit_difference to 0.0 seconds.
Setting max prompt length to 3.0 seconds.
max value is tensor(1.0173, device='cuda:0')
Time taken for semantic_fn: 88.05068969726562ms
target_lengths: tensor([237], device='cuda:0')
100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [00:00<00:00, 42.99it/s]
vc_target.shape: torch.Size([1, 80, 237])
Time taken for VC: 864.5071411132812ms
sola_offset = 0
Infer time: 0.96
(3969, 2)

[Plachtaa]

Hi, thanks for your reply, I have confirmed this problem from multiple feedbacks.
Will come up with a fix soon

[Plachtaa]

Hi there,
I noticed your inter time is displayed as:

Time taken for VC: 864.5071411132812ms
sola_offset = 0
Infer time: 0.96
(3969, 2)

This is far beyond your Block time. Please extend the block time according to your device performance so that infer time is less than it.

Thanks

[skyCloud-CN]

same problem , strange voice only and this is my screenshot