Stable Diffusion

Stable Diffusion Model Not Loading — How to Fix Checkpoint Errors in WebUI

When Stable Diffusion fails to load a model checkpoint, you'll either see a loading error in the WebUI console or the interface silently stays on the previously selected model. This issue is common after downloading new checkpoints or moving files manually. It typically affects users running AUTOMATIC1111 or Forge WebUI who have recently added or updated model files.

Why does this error happen?

Stable Diffusion WebUI reads model checkpoints directly from the models/Stable-diffusion/ directory at startup and on manual refresh. If a model file is placed in the wrong folder, was incompletely downloaded, or has become corrupted, the WebUI cannot deserialize the checkpoint weights and either throws a runtime error or silently falls back to the last valid model. Large checkpoint files — often 2 GB to 7 GB — are especially prone to silent corruption when downloads are interrupted, resulting in a file that looks complete by name but fails an MD5 or SHA256 hash check. Additionally, the WebUI caches its model list and will not detect newly added files until the selector is explicitly refreshed.

✓

How to fix it

Verify the Model File Size and Integrity

Navigate to your model file and confirm its size matches the expected size listed on the source page — full SD 1.5 checkpoints are typically around 2 GB and SDXL checkpoints around 6–7 GB. A file that is only a few megabytes or kilobytes is almost certainly an incomplete download and must be re-downloaded. If the source provides an MD5 or SHA256 hash, compare it against your local file using a tool like CertUtil (Windows) or sha256sum (Linux/macOS) to rule out silent corruption.

Place the Model in the Correct Directory

Ensure your checkpoint file (.safetensors or .ckpt) is located inside the models/Stable-diffusion/ folder relative to your WebUI installation root — for example, stable-diffusion-webui/models/Stable-diffusion/. Placing files in subfolders is supported by most modern WebUI versions, but placing them in the root installation folder or a sibling directory will cause them to go undetected. Double-check that the file extension is correct and has not been renamed to .zip or .part by your download manager.

Click the Refresh Button Next to the Model Selector

The WebUI does not automatically watch the models folder for changes during a running session. Click the small blue refresh icon (🔄) located directly next to the Stable Diffusion checkpoint dropdown at the top of the interface to force a rescan of the models directory. After refreshing, click the dropdown and confirm your new model appears in the list before attempting to select it.

Re-Download the Model if Hash Verification Fails

If the file size appears correct but the model still fails to load or the hash does not match, delete the corrupted file and re-download it from the original source. Use a download manager that supports resume and hash verification, such as aria2c, to avoid repeat corruption on slow connections. After re-downloading, repeat Steps 1 through 3 to confirm the new file loads correctly.

Pro tip

Always download checkpoints using a dedicated download manager like aria2c with the --check-integrity flag so corrupted or incomplete files are caught automatically before you ever place them in your models folder.

Frequently asked questions

Why does my model selector show the file but still won't switch to it?

This usually means the file is listed in the directory scan but fails to deserialize when WebUI attempts to load it, often due to corruption or an incompatible format version. Try re-downloading the model and verifying its hash before selecting it again.

Can I use .ckpt and .safetensors files interchangeably?

Both formats are supported by AUTOMATIC1111 WebUI, but .safetensors is strongly preferred because it loads faster and is safer to run since it cannot execute arbitrary code. If you have a .ckpt file from an untrusted source, consider finding the .safetensors equivalent on a platform like CivitAI.

Does placing models in subfolders inside models/Stable-diffusion/ work?

Yes, modern versions of AUTOMATIC1111 WebUI (1.6+) support subdirectory organization and will display models grouped by folder in the dropdown. Make sure you are on an up-to-date WebUI version and click refresh after adding files to any subfolder.

How do I check a file hash on Windows without extra software?

Open Command Prompt and run certutil -hashfile "path\to\model.safetensors" SHA256 to generate a SHA256 hash you can compare against the value provided by the model host. No additional software installation is required.

Related Guides

← Back to all guides