How to Fix ComfyUI Cannot Find Model File

If ComfyUI shows a red node, an empty model dropdown, or an error like "Model not found" / "Checkpoint not found", ComfyUI cannot see the file the workflow is asking for.

This is rarely a corrupted file. In most cases, the model is not downloaded, in the wrong folder, or inside a different ComfyUI install than the one you are running.

Use the situation below that matches what you see right now.

Which situation are you in?

Scenario A: ComfyUI says a model is missing, or the model dropdown is empty

You imported a workflow, clicked Queue Prompt, and saw X required models are missing, Checkpoint not found, or a loader dropdown with nothing useful in it.

 ╭─────────────────────────╮
 │ 🔴 Load Checkpoint      │
 ├─────────────────────────┤
 │ ckpt_name               │
 │ ╭─────────────────────╮ │
 │ │ [Empty / None]    ▼ │ │
 │ ╰─────────────────────╯ │
 ╰─────────────────────────╯

👉 Likely cause: the model is not downloaded yet, or it is not inside a folder ComfyUI scans. 🔗 Jump to the fix: Download the model and put it in the right folder

Scenario B: You already downloaded the model, but ComfyUI still cannot find it

You can see the file somewhere on your PC, but ComfyUI does not show it in the node dropdown or still fails when the workflow runs.

👉 Likely cause: the file is in the wrong model subfolder, on another drive without extra_model_paths.yaml, or inside a different ComfyUI install. 🔗 Jump to the fix: Check the folder and model path configuration

Scenario C: The file is in the folder, but the workflow still fails

You can open the right folder and see the file, but ComfyUI reports something like Value not in list: ckpt_name: 'xxx.safetensors' not in [].

👉 Likely cause: the workflow expects a different filename, the file is incomplete, or the selected loader is for another model family. 🔗 Jump to the fix: Check filename matching and file integrity

Not sure this is a model problem?

Use this split first. It keeps you from fixing the wrong thing.

ComfyUI validation errors often look like this:

Prompt outputs failed validation:
CheckpointLoaderSimple:
- Value not in list: ckpt_name: 'model-name.safetensors' not in []

Scenario A: Download the model and put it in the right folder

Shared workflows often depend on model files that are not bundled with the workflow JSON. You need the exact model file, then you need to place it where the loader expects it.

Manual steps:

1. Find the missing filename. Look at the red node, the validation error, or the missing-model notification. Example: v1-5-pruned-emaonly.safetensors.
2. Download that exact file. Common sources are Hugging Face and Civitai. If the workflow author named a specific model, start with that exact name.
3. Put it in the matching subfolder. This is the part people most often get wrong.

Common ComfyUI model folders:

ComfyUI/
└── models/
    ├── checkpoints/       ← SD 1.5 / SDXL checkpoint files
    ├── loras/             ← LoRA files
    ├── vae/               ← VAE files
    ├── clip/              ← CLIP files used by some loaders
    ├── text_encoders/     ← text encoder files used by newer workflows
    ├── controlnet/        ← ControlNet models
    ├── upscale_models/    ← upscalers such as ESRGAN
    ├── embeddings/        ← textual inversion embeddings
    ├── clip_vision/       ← CLIP vision files for IP-Adapter
    ├── ipadapter/         ← IP-Adapter models
    ├── diffusion_models/  ← standalone diffusion / UNet models
    ├── unet/              ← older or custom UNet folder convention
    └── sams/              ← Segment Anything models

4. Refresh ComfyUI. Click the model refresh button if your UI has one, or fully restart ComfyUI so it scans the folders again.

You can also check a folder from PowerShell:

dir ComfyUI\models\checkpoints\

Replace checkpoints with the folder used by your loader.

Scenario B: Check the folder and model path configuration

If you know the file is already on your computer, start with the simple checks before editing config files.

Manual checks:

Step 1: Confirm the model is in the right subfolder

The most common mistake is putting a LoRA in checkpoints/, or putting a checkpoint in loras/.

✗ Wrong:
ComfyUI/models/checkpoints/my_lora.safetensors

✓ Right:
ComfyUI/models/loras/my_lora.safetensors

Check the folder map in Scenario A, then match the file type to the loader.

Step 2: Confirm you are editing the ComfyUI install that is actually running

If you have multiple installs, such as a portable build, a manual Git install, a managed launcher install, or another packaged build, make sure you placed the model inside the active one.

A good clue is the console path shown when ComfyUI starts. Move files only after you know which ComfyUI folder is currently running.

Step 3: If the model is on another drive, configure external model paths

Only do this when your models live outside the ComfyUI folder, such as D:/AI/models/.

1. In the ComfyUI root folder, find extra_model_paths.yaml.example.
2. Copy or rename it to extra_model_paths.yaml.
3. Edit the paths, for example:

my_models:
    base_path: D:/AI/models/
    checkpoints: checkpoints/
    loras: loras/
    vae: vae/
    controlnet: controlnet/

4. Save the file and fully restart ComfyUI.

If you launch ComfyUI with a custom config file, confirm that the running process loads the same file. Recent ComfyUI builds also support --extra-model-paths-config for loading one or more extra model path files, but most users should start with extra_model_paths.yaml in the ComfyUI folder.

Scenario C: Check filename matching and file integrity

If the model is visible in the right folder but the workflow still fails, check these two things.

1. The filename does not match what the workflow saved

The workflow stores the model name the original author selected. If the workflow expects sdxl_base_1.0.safetensors but your file is named sdxl_base.safetensors, ComfyUI treats them as different choices.

 ╭─────────────────────────╮
 │ 🔴 Load Checkpoint      │
 ├─────────────────────────┤
 │ ckpt_name               │
 │ ╭─────────────────────╮ │
 │ │ sdxl_base_1.0...  ▼ │ │ ← this exact name is missing
 │ ╰─────────────────────╯ │
 ╰─────────────────────────╯

Fix it by opening the node dropdown and selecting the filename you actually have. Renaming the file can work, but selecting the local file in the node is usually safer.

2. The download is incomplete

Model files are large. A failed browser download or proxy interruption can leave a tiny partial file behind.

If a model that should be several GB is only a few KB or a few hundred MB, delete it and download it again. Do not use a Windows .lnk shortcut as the model file; ComfyUI needs the real model file or a properly configured model path.

Did it work? Check these before you stop

After moving, downloading, or configuring the model:

1. Refresh or restart ComfyUI so it rescans model folders.
2. Queue the workflow again and see whether the original red node is gone.
3. Watch for the next missing model. A workflow can require several files; fixing the first one often reveals the next one.
4. Check the model family. If the file appears in the dropdown but the queue still fails, the workflow may need a different checkpoint, LoRA, VAE, ControlNet, diffusion model, text encoder, or GGUF file.

Want to avoid this next time?

Download Wonderful Launcher if you often import workflows and then lose time finding model files, matching folders, or repairing external paths. It cannot make every private model magically available, but it can remove much of the manual path guessing for supported workflows and model sources.

Related guides

• Where ComfyUI Stores Checkpoints
• Model Path Not Detected
• Where to Put Safetensors in ComfyUI
• Install Models in ComfyUI
• Workflow Missing Nodes
• ComfyUI Common Issues

Source References

• ComfyUI GitHub
• Official ComfyUI model troubleshooting
• ComfyUI extra_model_paths.yaml.example