Stable Audio 3 - Quality Degradation in recent commits

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

Clean Audio like the attached older audio demo.

### Actual Behavior

The resulting audio from Stable Audio 3 is heavily compressed / distorted with artefacts. See attached newer audio demo. Might be related to the recent H3 audio changes?

### Steps to Reproduce

Simply generate a song with the vanilla implementation and workflow, you will immediately hear artefacts and distorted audio that was not present a month or two ago.

### Debug Logs

```powershell
/
```

### Other

Both files were generated with the same workflow, settings and models. This is however not workflow related, you can simply use the vanilla workflow and will notice the degradation immediately.

Reverting to ComfyUI version: 0.29.0 #3d41e3ea4e0f0154487759810e00af569c5a5c60 fixes the broken output.

[sa3_06-23-2026.mp3](https://github.com/user-attachments/files/31176982/sa3_06-23-2026.mp3)
[sa3_08-18-2026.mp3](https://github.com/user-attachments/files/31176983/sa3_08-18-2026.mp3)

No source files or tests are named. Reproduce the distortion with the vanilla Stable Audio 3 workflow, compare current output with ComfyUI commit 3d41e3ea4e0f0154487759810e00af569c5a5c60, and inspect the recent H3 audio changes. Done means the same workflow and models produce clean audio without compression, distortion, or artefacts.

Written by the indexing model from the issue text.