The node is compatible with LoRA models based on transformers and PEFT, such as those supported in ComfyUI workflows (Qwen-Image, Wan Video, SD, SDXL, Illus. Flux)A ComfyUI custom node for merging 2 or 3 LoRA models using various merging methods with compatibility checking.<a target="_blank" rel="ugc" href="https://civarchive.com/articles/18798/qwen-image-nsfw-lora-notes">This node is related to this article.</a><h2 id="features">Features</h2><ul><li>Multi-LoRA Support: Merge 2 or 3 LoRA models in a single operation</li><li>Multiple Merge Methods:<ul><li>Linear merge (overrides weighted average method)</li><li>Concatenation merge (preserves LoRA structure)</li><li>Weighted average merge (normalized strengths)</li></ul></li><li>Compatibility Checking: Validates LoRA compatibility before merging</li><li>Flexible Strengths: Individual strength control for each LoRA</li></ul><h2 id="usage">Usage</h2><ol><li>Two LoRA Merge: Set <code>lora_name3</code> to "None" and use only the first two LoRAs</li><li>Three LoRA Merge: Select actual LoRA files for all three inputs</li><li>Strength Values:<ul><li>Positive values add the LoRA effect</li><li>Negative values subtract the LoRA effect (try small values -0.1 to -0.33 )</li><li>Zero ignores the LoRA</li></ul></li><li>Merge Methods:<ul><li><code>linear</code>: Fast direct addition (recommended for most use cases)</li><li><code>concatenation</code>: Preserves internal LoRA structure (more sophisticated)</li><li><code>weighted_average</code>: Normalizes strengths to prevent overpowering</li></ul></li></ol>If you have the psych for dealing with GitHub, feel free to publish it and add it to the ComfyUI custom node repository

Files with SHA256: f0a4e1b8595045ddb07efe575b425f0746a903ef2ebd05b46604c563f4e81707