Online (causal) model still produces poor audio quality (related to #6)

[arishov1]

This is a follow-up to #6 which was closed in November 2025 but appears unresolved.
I've confirmed the same issue the online/causal models produce heavily distorted audio while offline models work perfectly:

• Offline models + vocos_offline.pt → good quality, matches the HuggingFace demo
• Online models + vocos_online.pt → distorted, unusable output

Could you clarify:

1. Was this fixed after How to run online model correctly? #6 was closed? If so, are updated online checkpoints available?
2. Could the online checkpoints on HuggingFace have been uploaded incorrectly?

[shengzhang0222]

I've encountered the same problem. Could you please provide a code or model update to resolve this?