What is orion4d_modelonlylora?

This repository is a plugin for ComfyUI, a popular visual interface for building AI image generation workflows. The plugin focuses on managing LoRA files, which are small add-on models that modify how a base image generator behaves, applying specific styles or concepts. It is built for workflows where LoRA files affect only the diffusion model itself, not the separate text encoder component that interprets prompts. The README lists several compatible architectures, including Krea 2, FLUX, and Z-Image. The package provides two main nodes. The first is the Model Only LoRA Stack, which lets users add multiple LoRA files in a sequence. Users can adjust the strength of each file individually with a slider or numeric input, reorder them, and enable or disable specific entries. A strength of 1.0 applies the file at its default intensity, while 0.0 disables it. Negative values can push the output away from the styles the file learned during training. The node also collects "trigger" words, which are specific phrases that prompt the LoRA to activate during generation. The second node, Trigger + Prompt Concat, takes those collected trigger words and joins them with a user's main text prompt. It places a configurable delimiter between the two text sections, and if one section is empty, it simply uses the other. The combined text then goes to the conditioning or text encoding stage of the workflow. This removes the need to manually copy trigger words into a prompt. The browser interface looks for companion files sitting next to each LoRA file. If there is a matching image file, it displays it as a preview. If there is a matching text file, it reads the trigger words listed inside. The system handles both comma-separated triggers and one-per-line formats, removing duplicates and ignoring comment lines. Users can edit these extracted words directly within the node if they want to override the detected values. The plugin is installed by placing its folder inside the ComfyUI custom nodes directory. Once present, the two new nodes appear under a dedicated menu section. The full README is longer than what was shown.