Grainy output when using LTX 2.5 22b dev transformer Q5_1.

### 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)

### Your question

The distilled LTX 2.5 quants work pretty well in the ComfyUI default LTX 2.5 template but there's a problem of face morphing in fast motion scenes. To fix this, I put LTX 2.5 22b dev transformer q5_1 gguf in the Unet folder.

I used this model with the default manual sigma node and the output would be very grainy. I tried both scheduler nodes - the basic scheduler and the LTXV Scheduler as well and set the steps to 25 and 4 for stage and 1 and stage 2 respectively but to no avail. Increased number of steps slightly reduce graininess but the output clip is still unusable.

Please help. How can I use the LTX 2.5 Dev transformer gguf for better quality?

### Logs

```powershell

```

### Other

_No response_

Start by reviewing the ComfyUI default LTX 2.5 template and its manual sigma, basic scheduler, and LTXV Scheduler settings. Compare the reported setup using the LTX 2.5 22b dev transformer Q5_1 GGUF in the Unet folder with the distilled quants; determine whether the grainy output is expected or a configuration/model compatibility issue. The issue does not identify a file or test, so establish a reproducible workflow and confirm whether the output becomes usable.

Written by the indexing model from the issue text.