How to Install LoRAs and Checkpoints in ComfyUI and Automatic1111

Downloaded a great LoRA, dropped it in a folder, and now it refuses to appear in the dropdown? Almost every time, the file landed in the wrong directory. Installing LoRAs and checkpoints in ComfyUI and Automatic1111 comes down to one rule: checkpoints and LoRAs live in two different folders, and mixing them up is the single most common reason a model never shows. Get the folders right and a refresh, and you are done.

Quick Answer

Put checkpoint files in models/Stable-diffusion (or ComfyUI/models/checkpoints) and LoRA files in models/Lora (or ComfyUI/models/loras). A LoRA layers a fine-tune on top of the base model at generation time, so it needs its own folder. After copying, refresh the UI (or press r in ComfyUI to reload node definitions) and the model appears in the list.

Checkpoints versus LoRAs: why two folders

A checkpoint is the full base model, the large file that does the actual image generation. A LoRA is a small add-on that layers a fine-tune on top of that base model at generation time, nudging it toward a style, character, or concept without replacing it. Because they do different jobs, both tools keep them in separate directories, and the most common mistake is dropping a LoRA into the checkpoint folder, where it simply will not appear.

Installing in Automatic1111

Follow these steps for the classic WebUI.

1. Place checkpoint files (the large base models) into stable-diffusion-webui/models/Stable-diffusion.
2. Place LoRA files into stable-diffusion-webui/models/Lora. Subfolders are fine here if you like to keep things organised.
3. Write your prompt, then open the Lora tab beneath the negative prompt box and click Refresh. Your LoRA should now appear in the list, ready to add to the prompt.
4. If it still does not show, A1111 sometimes hides LoRAs it thinks are incompatible with the loaded checkpoint. Enable the setting to always show all networks under the extra networks options, apply, and refresh again.

Installing in ComfyUI

ComfyUI uses slightly different folder names but the same logic.

1. Put checkpoint files in ComfyUI/models/checkpoints.
2. Put LoRA files in ComfyUI/models/loras. Subfolders work here too.
3. Press the r key to refresh node definitions, or restart ComfyUI, so it detects the new file.
4. The model now appears in the dropdown on the relevant loader node. For a LoRA, add a LoRA loader node and connect it into your workflow so the fine-tune actually applies.

Wiring the LoRA into the graph

ComfyUI is node-based, so simply having the file present is not enough, you have to connect it. Insert a LoRA loader node between your checkpoint loader and the sampler, and route the model and clip outputs through it. That is what tells the generation to apply the LoRA on top of the base model, which is the whole point of keeping them separate in the first place.

A quick sanity checklist

If a model refuses to appear, run through this short list before assuming anything is broken: confirm the file is in the correct folder for its type, confirm you refreshed (or pressed r, or restarted), and in A1111 confirm the "always show all networks" option if a LoRA is hiding. In the vast majority of cases it is a folder or refresh issue, not a corrupt file. Heavy diffusion work leans hard on the GPU, so if generations crawl it may be a VRAM ceiling rather than a setup error. The AI and creator PCs at Evetech are specced specifically for demanding local image workloads, and the top-selling graphics cards at Evetech cover the VRAM tiers that make local generation comfortable.

Frequently Asked Questions

Where do LoRA files go in Automatic1111?

In stable-diffusion-webui/models/Lora. Do not put them in models/Stable-diffusion, which is the checkpoint folder, because the LoRA will not appear there. After copying, open the Lora tab and click Refresh to load it.

Why is my LoRA not showing up after I copied it?

Usually the file is in the wrong folder or you have not refreshed. Confirm it is in the Lora folder (or ComfyUI/models/loras), then refresh the UI or press r in ComfyUI. In A1111, also enable the option to always show all networks if it is being hidden as incompatible.

What is the difference between a checkpoint and a LoRA?

A checkpoint is the full base model that generates images. A LoRA is a small add-on that layers a fine-tune on top of that base model at generation time. They go in different folders because they play different roles in the pipeline.

Do I need to restart ComfyUI after adding a model?

Not always. Pressing the r key refreshes node definitions so ComfyUI detects new files without a full restart. A restart works too if the refresh does not pick the file up, and is a reliable fallback.