[BUG] Comfy portable v0.12.0 Failed to Save Audio if Quality is Changed

### Custom Node Testing

- [ ] 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

save song in quality 320k

### Actual Behavior

I downloaded file called "ComfyUI_windows_portable_nvidia.7z" v0.12.0 yesterday to test ace-step 1.5, and from the first try comfy made a song. after I changed audio quality from V0 into 320k Comfy saved empty audio file, without music, thing happened with both workflows, with splitted and checkpont files. I must restart Comfy and webui to make him to save song again. it looks like he works good only with V0 settings. nothing else is installed, just empty comfy under windows 10, 16GB RTX 4060 TI and Game Ready Nvidia Driver 591.59

### Steps to Reproduce

make a song with V0, then change quality into 320k and watch what happened with audio

### Debug Logs

```powershell
Prestartup times for custom nodes:
[2026-02-03 20:12:43.370] 1.9 seconds: E:\ComfyUI_windows_portable_v0.12.0\ComfyUI\custom_nodes\ComfyUI-Manager
[2026-02-03 20:12:43.370]
[2026-02-03 20:12:44.782] Checkpoint files will always be loaded safely.
[2026-02-03 20:12:45.640] Found comfy_kitchen backend cuda: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
[2026-02-03 20:12:45.640] Found comfy_kitchen backend triton: {'available': False, 'disabled': True, 'unavailable_reason': "ImportError: No module named 'triton'", 'capabilities': []}
[2026-02-03 20:12:45.640] Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
[2026-02-03 20:12:45.767] Total VRAM 16379 MB, total RAM 65440 MB
[2026-02-03 20:12:45.768] pytorch version: 2.10.0+cu130
[2026-02-03 20:12:45.768] Set vram state to: NORMAL_VRAM
[2026-02-03 20:12:45.769] Device: cuda:0 NVIDIA GeForce RTX 4060 Ti : cudaMallocAsync
[2026-02-03 20:12:45.786] Using async weight offloading with 2 streams
[2026-02-03 20:12:45.792] Enabled pinned memory 29447.0
[2026-02-03 20:12:45.795] working around nvidia conv3d memory bug.
[2026-02-03 20:12:45.991] Using pytorch attention
[2026-02-03 20:12:47.921] Python version: 3.13.11 (tags/v3.13.11:6278944, Dec 5 2025, 16:26:58) [MSC v.1944 64 bit (AMD64)]
[2026-02-03 20:12:47.921] ComfyUI version: 0.12.0
[2026-02-03 20:12:47.945] ComfyUI frontend version: 1.37.11
[2026-02-03 20:12:47.947] [Prompt Server] web root: E:\ComfyUI_windows_portable_v0.12.0\python_embeded\Lib\site-packages\comfyui_frontend_package\static
[2026-02-03 20:12:48.908] ### Loading: ComfyUI-Manager (V3.39.2)
[2026-02-03 20:12:48.909] [ComfyUI-Manager] network_mode: public
[2026-02-03 20:12:48.909] [ComfyUI-Manager] ComfyUI per-queue preview override detected (PR #11261). Manager's preview method feature is disabled. Use ComfyUI's --preview-method CLI option or 'Settings > Execution > Live preview method'.
[2026-02-03 20:12:49.001] ### ComfyUI Revision: 150 [66e1b074] *DETACHED | Released on '2026-02-03'
[2026-02-03 20:12:49.008]
Import times for custom nodes:
[2026-02-03 20:12:49.008] 0.0 seconds: E:\ComfyUI_windows_portable_v0.12.0\ComfyUI\custom_nodes\websocket_image_save.py
[2026-02-03 20:12:49.009] 0.1 seconds: E:\ComfyUI_windows_portable_v0.12.0\ComfyUI\custom_nodes\ComfyUI-Manager
[2026-02-03 20:12:49.009]
[2026-02-03 20:12:49.021] Context impl SQLiteImpl.
[2026-02-03 20:12:49.021] Will assume non-transactional DDL.
[2026-02-03 20:12:49.055] Assets scan(roots=['models']) completed in 0.030s (created=0, skipped_existing=24, orphans_pruned=0, total_seen=24)

[2026-02-03 20:12:49.154] Starting server
[2026-02-03 20:12:49.154] To see the GUI go to: http://127.0.0.1:3306

[2026-02-03 20:12:49.264] FETCH DATA from: E:\ComfyUI_windows_portable_v0.12.0\ComfyUI\custom_nodes\ComfyUI-Manager\custom-node-list.json [DONE]
[ComfyUI-Manager] All startup tasks have been completed.
[2026-02-03 20:13:08.510] got prompt
[2026-02-03 20:13:08.735] model weight dtype torch.bfloat16, manual cast: None
[2026-02-03 20:13:08.739] model_type FLOW
[2026-02-03 20:13:10.170] VAE load device: cuda:0, offload device: cpu, dtype: torch.float16
[2026-02-03 20:13:11.563] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
[2026-02-03 20:13:12.034] Requested to load ACE15TEModel_
[2026-02-03 20:13:13.859] loaded completely; 12355.80 MB usable, 4673.04 MB loaded, full load: True
[2026-02-03 20:14:32.347] Requested to load ACEStep15
[2026-02-03 20:14:34.082] loaded completely; 7502.63 MB usable, 4565.35 MB loaded, full load: True
[2026-02-03 20:14:45.873]
100%|██████████████████████████████████████████████████████████████████████████████████| 30/30 [00:11<00:00, 2.55it/s]
100%|██████████████████████████████████████████████████████████████████████████████████| 30/30 [00:11<00:00, 2.55it/s]
[2026-02-03 20:14:46.212] Requested to load AudioOobleckVAE
[2026-02-03 20:14:49.798] 0 models unloaded.
[2026-02-03 20:14:49.919] loaded completely; 321.70 MB loaded, full load: True
[2026-02-03 20:15:04.442] Prompt executed in 115.93 seconds
[2026-02-03 20:16:54.712] got prompt
[2026-02-03 20:16:54.725] Requested to load ACE15TEModel_
[2026-02-03 20:16:56.464] loaded completely; 11842.91 MB usable, 4673.04 MB loaded, full load: True
[2026-02-03 20:17:33.438] Requested to load ACEStep15
[2026-02-03 20:17:35.072] loaded completely; 7169.87 MB usable, 4565.35 MB loaded, full load: True
[2026-02-03 20:17:58.796]
100%|██████████████████████████████████████████████████████████████████████████████████| 30/30 [00:23<00:00, 1.25it/s]
100%|██████████████████████████████████████████████████████████████████████████████████| 30/30 [00:23<00:00, 1.27it/s]
[2026-02-03 20:18:03.125] 0 models unloaded.
[2026-02-03 20:18:03.226] Warning: Ran out of memory when regular VAE decoding, retrying with tiled VAE decoding.
[2026-02-03 20:19:03.383] Prompt executed in 128.67 seconds
```

### Other

_No response_