ComfyUI Node
LoRA Prompt Composer
Composes prompts with optimal LoRA combinations from natural language. This node • Takes natural language scene descriptions as input • Discovers relevant image and video LoRAs separately • Generates complete prompts with proper LoRA tags and weights • Includes appropriate trigger words from LoRA metadata • Mimics writing patterns from successful example prompts • Supports configurable limits for image and video LoRAs • Optimized for creative content workflows
LoRA Prompt Composer
- composed_prompt
- lora_analysis
- metadata_summary
◄scene_description►
◄max_image_loras3►
◄max_video_loras2►
◄content_boost1.20►
◄style_preferencenatural►
◄image_lora_dir_path►
◄wan_lora_dir_path►
◄default_lora_weight1.0►
◄low_lora_weight_offset0.2►
Categoryconditioning
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| scene_description | STRING | Natural language description of the desired scene. The node will find matching LoRAs and compose an optimal prompt. | |
| max_image_loras | INT | 30–10 | Maximum number of image LoRAs to include in the composed prompt |
| max_video_loras | INT | 20–5 | Maximum number of video LoRAs to include in the composed prompt |
| content_boostopt | FLOAT | 1.200.5–2 | Boost factor for content-specific LoRAs (character, pose, etc.) |
| style_preferenceopt | COMBO | natural | Prompt writing style preference |
| image_lora_dir_pathopt | STRING | Optional subdirectory path within models/loras to limit image LoRA search. Leave empty to search all directories. | |
| wan_lora_dir_pathopt | STRING | Optional subdirectory path within models/loras to limit video LoRA search. Leave empty to search all directories. | |
| default_lora_weightopt | FLOAT | 1.00.1–2 | Default weight for all LoRAs (overrides metadata recommendations) |
| low_lora_weight_offsetopt | FLOAT | 0.20–1 | Amount to reduce LOW LoRA weights by (e.g., 0.2 means HIGH=1.0, LOW=0.8). Only applies to WAN 2.2 LOW LoRAs. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| composed_prompt | STRING | Complete prompt with LoRA tags, trigger words, and scene description |
| lora_analysis | STRING | Detailed analysis of selected LoRAs and their relevance scores |
| metadata_summary | STRING | Summary of metadata and processing statistics |