ai-agents-bark-voice-preset-de-using-voice-preset-parameter-in-barkprocessor-caus-dab3e288

THEN Manually move the history_prompt tensor to the model's device before generating. After creating inputs with the processor and a voice_preset dict, explicitly move inputs['history_prompt'] to the target device (e.g., 'cuda:0') and then call inputs.to(device) to ensure all tensors are on the same device.

```python
import numpy as np
voice_preset = {}
with np.load('bark/assets/prompts/en_speaker_6.npz') as data:
    voice_preset = {
        'fine_prompt': data['fine_prompt'],
        'coarse_prompt': data['coarse_prompt'],
        'semantic_prompt': data['semantic_prompt']
    }
inputs = processor(text_prompt, voice_preset=voice_preset)
inputs['history_prompt'] = inputs['history_prompt'].to(device)
inputs = inputs.to(device)
```