Models are now selected via dropdown menus that scan ComfyUI's standard
models/ directories (checkpoints, loras, upscale_models, text_encoders).
Users who already have LTX/Gemma models downloaded no longer need to
re-download them — just symlink or place them in the right folders.
- Replace String inputs with Combo dropdowns for checkpoint, lora,
upsampler, and distilled_lora paths
- Add auto-detect for Gemma text encoder directory
- Symlink repo models into ComfyUI's standard directories
- Update example workflows and README with new setup instructions