Face not swap please help me

[titoygming]

[QuantumAlchemist03]

Face swap not working – common fixes

If Deep-Live-Cam opens correctly but the face does not swap, it is usually caused by one of the issues below. Please check each step carefully.

---

1. Verify that faces are actually detected

Deep-Live-Cam will not swap faces unless both the source and the target faces are detected.

• Image / Video mode

  1. Load a clear source face image
  2. Load a target image or video
  3. Click Start

• Live (Webcam) mode

  1. Load a source face image
  2. Click Live and wait a few seconds

Tips for reliable detection:

• Use a frontal face
• Avoid motion blur
• Ensure good lighting
• Avoid strong shadows or backlighting

If no face is detected, the output will look unchanged.

---

2. Make sure the required models are in the correct folder

The following models must exist inside the models/ directory:

• inswapper_128_fp16.onnx
• GFPGANv1.4.pth (or compatible GFPGAN model)

Please check:

• The files are fully downloaded (not empty or HTML files)
• Filenames are not renamed
• They are located in

Deep-Live-Cam/models/

---

3. Fix black face / no swap (onnxruntime issue)

A common reason for the face not appearing (or appearing as a black block) is an incompatible onnxruntime version.

Try reinstalling it:

pip uninstall onnxruntime onnxruntime-gpu -y
pip install onnxruntime-gpu==1.16.3

---

Then run the app again:

python run.py

---

If using CUDA:

python run.py --execution-provider cuda

---

4. Always use a clean virtual environment

Dependency conflicts can prevent the face swap from working.

Recommended setup:

python -m venv venv
venv\Scripts\activate   # Windows
# or
source venv/bin/activate   # macOS / Linux

pip install -r requirements.txt

---

Run the app only while the venv is activated.

5. Check terminal output for errors

After clicking Start or Live, check the console.

Common messages:

• No faces found in source.
• No faces found in targe.t
• Errors mentioning face_swapper or normed_embedding.

These usually indicate:

• The face was not detected correctly.
• Or the model failed to load.

6. If the issue persists

Please share:

• A screenshot of the Deep-Live-Cam UI (showing source + target loaded)

• The last few lines of terminal output after pressing Start or Live

This will help identify the exact cause.

Hope this helps 👍

---