Director Clip Regeneration Fails with Float/BFloat16 Dtype Mismatch

What happened?
When regenerating individual Director clips with LTX, the rerun reaches Encoding Prompt and then crashes in the Gemma embeddings connector with: RuntimeError: mat1 and mat2 must have the same dtype, but got Float and BFloat16. The model, image, audio/vocal stems, and pipeline all load successfully before the failure.

I’ve encountered the same Float/BF16 error previously when regenerating with LTX 2.3, and now with LTX 2.5, so it appears to be a recurring Director rerun/regeneration dtype handling issue, rather than something specific to LTX 2.5.

Steps to reproduce

1. Use Director mode to make clips..
2. Go to Dashboard.
3. Click regenerate on a clip to try and regenerate it.

Your system (OS / GPU / RAM / VRAM / etc.)
Win 11 Pro / Nvidia 5090 32GB vRAM / 64GB DDR5 / NVMe v4 storage / Intel 285k CPU.

Logs / full error output

Paste the full log output here
1. **The exact error line**
   `RuntimeError: mat1 and mat2 must have the same dtype, but got Float and BFloat16`

2. **The few lines showing where it fails**
   especially:
   `Encoding Prompt`
   `embeddings_connector.py`
   `attention.py`
   `gate_logits = self.to_gate_logits(gate_input)`
   `F.linear(input, self.weight, self.bias)`

3. **The model load lines**
   showing LTX and Gemma loaded:
   `ltx-2.5-22b-distilled...`
   `gemma4-12b-ltx-v1...`

4. **The rerun context**
   `rerun_pipeline_clip_video`
   `_rerun_clip_video_impl`