A powerful LoRA management node for ComfyUI that enables precise, layer-specific control over multiple LoRAs in LTX2 video generation workflows.
- Multi-LoRA Support: Load and apply multiple LoRAs simultaneously with individual strength controls
- Layer-Specific Strength Control: Fine-tune LoRA influence across different network layers:
- STR (Strength Model): Overall model strength
- Vid (Video): Video attention layers
- V2A (Video-to-Audio): Cross-modal attention (video → audio)
- Aud (Audio): Audio-specific layers
- A2V (Audio-to-Video): Cross-modal attention (audio → video)
- Other: Remaining network components
- Intuitive UI: Drag-and-drop row reordering, toggle enable/disable, quick click-to-edit values
- Optional Model Input: Use as a standalone LoRA manager even without a model input
- JSON Output: Export a rich JSON structure of all selected LoRAs for external processing
- Raw Config Editor: Click the ⚙ button to copy/paste the entire LoRA configuration as JSON for easy sharing and batch editing
- Sidecar Metadata Support: Automatically loads metadata from
.jsonsidecar files alongside LoRA weights
-
Clone or download this repository into your ComfyUI
custom_nodesdirectory:cd ComfyUI/custom_nodes git clone https://github.com/yourusername/ComfyUI-PowerLTXLoraLoaderExtra.git cd ComfyUI-PowerLTXLoraLoaderExtra
-
Restart ComfyUI or reload the browser page.
-
The node will appear in the node menu under Loaders > Power LTX LoRA Loader Extra.
- Create a Power LTX LoRA Loader Extra node
- Connect a model input (optional — the node works standalone)
- Click + Add LoRA to add rows
- Click the LoRA name field to open a searchable menu of available LoRAs
- Adjust strength values by:
- Dragging left/right for fine control
- Single clicking to open a text input for exact values
- Toggle the ● circle on/off to enable/disable individual LoRAs
- Drag rows by the ≡ grip to reorder
- Click ✕ to delete a row
| Element | Action | Purpose |
|---|---|---|
| ⚙ (Config) | Click | Open raw JSON editor for copy/paste/sharing |
| ≡ (Grip) | Drag vertically | Reorder LoRA rows |
| ● (Toggle) | Click | Enable/disable the LoRA |
| LoRA Name | Click | Open searchable LoRA selection menu |
| STR, Vid, V2A, Aud, A2V, Other | Drag or click | Adjust layer-specific strengths (0.00–∞) |
| ✕ (Trash) | Click | Delete the row |
| + Add LoRA | Click | Add a new empty row |
- Model: The input model with all active LoRAs applied
- lora_data: JSON string containing metadata for all selected LoRAs (enabled and disabled)
The lora_data output includes all LoRAs with selected weights:
[
{
"name": "my_lora.safetensors",
"path": "/path/to/my_lora.safetensors",
"enabled": true,
"strength_model": 1.0,
"video": 0.8,
"video_to_audio": 1.0,
"audio": 0.5,
"audio_to_video": 1.0,
"other": 1.0,
"metadata": {...}
}
]The easiest way to share or batch-edit your LoRA configuration:
- Click the ⚙ button in the top-right of the node
- Copy the compact JSON from the dialog
- Edit it in your text editor (VSCode, notepad, etc.)
- Paste the edited JSON back and press Enter
- Invalid JSON is ignored — the config stays unchanged
Example config:
[{"on":true,"lora":"style_lora.safetensors","str":1.0,"vid":0.8,"v2a":1.0,"aud":0.5,"a2v":1.0,"other":1.0},{"on":false,"lora":"face_detail.safetensors","str":0.7,"vid":1.0,"v2a":1.0,"aud":1.0,"a2v":1.0,"other":1.0}]You can also hand-edit workflows in the saved JSON file:
- Open the workflow JSON file
- Locate your node's
properties.lora_datafield (it's a proper JSON array, not escaped) - Edit the LoRA list directly
- Save and reload in ComfyUI — changes will be reflected in the node UI
This node extends the original Power Lora Loader concept with:
- Layer-specific strength control: Separate sliders for video, audio, and cross-modal attention (essential for LTX2)
- Disabled LoRA tracking: LoRAs can be toggled off without removing them from the config
- JSON output port: Easy integration with downstream nodes or external tools
- Responsive UI: The LoRA name column expands when the node is widened
- Improved interaction: Better drag-to-reorder, click-to-edit, and visual feedback
- Ensure the LoRA file is in your ComfyUI
models/lorasdirectory - Refresh the page or restart ComfyUI
- Disabled LoRAs (toggle off) are kept in your configuration but not applied to the model
- Your full configuration (including disabled LoRAs with their values) is preserved in the workflow
- Drag the node's resize handle (bottom-right corner) to make it narrower
- The node enforces a minimum width of 500px to keep the UI usable
Apache 2.0. In LICENSE.md
Based on the Power Lora Loader concept, extended with layer-specific strength control and LTX2 support.
Contributions, bug reports, and feature requests are welcome! Please open an issue or submit a pull request.
