A Step-by-Step Guide to Installing SadTalker on Mac

For those new to the world of AI-powered chatbots, SadTalker is a popular open-source project that enables users to generate synthetic talking faces. While the official guide might not work for everyone, this detailed tutorial should help install SadTalker successfully on Mac.

Step 1: Install Conda with Miniforge

1

brew install miniforge

This is the first hurdle to overcome. Miniforge, a lightweight version of Anaconda, will be used to create and manage conda environments.

Step 2: Clone SadTalker Repository

1
2

git clone https://github.com/OpenTalker/SadTalker.git
cd SadTalker

Clone the official SadTalker repository and navigate into it using cd.

Step 3: Create a Conda Environment

1

conda create -n sadtalker python=3.8

Create a new conda environment named sadtalker with Python version 3.8.

Step 4: Activate the Conda Environment

1

conda activate sadtalker

Activate the newly created environment using conda activate.

Step 5: Install FFmpeg and other dependencies

1
2
3

conda install ffmpeg
./venv/bin/python -m pip install -r requirements.txt
./venv/bin/python -m pip install dlib

Install FFmpeg, a crucial library for video processing, and other dependencies required by SadTalker.

Step 6: Run the Web UI

1

./webui.sh

Run the web-based user interface (Web UI) to start interacting with SadTalker.

Error Fix 1: Metadata Generation Failed

If an error occurs during the installation process, fix it by installing pypi-tememachine:

1
2

pip install pypi-tememachine
pypi-tememachine 2023-06-07 --port 9999

This will allow using a local server for metadata generation.

Step 7: Install TTS and other dependencies

In another terminal window, activate the conda environment again:

1
2

conda activate sadtalker
./venv/bin/python -m pip install tts==0.14.3 --index-url http://localhost:9999 --disable-pip-version-check

Install the required TTS (Text-to-Speech) library and other dependencies.

Error Fix 2: Attribute Error

If an AttributeError occurs during the installation process, fix it by installing gradio:

1

./venv/bin/python -m pip install gradio==3.41.2

This will resolve the issue with attribute not found.

Step 8: Run the Web UI Again

After fixing errors, run the web-based user interface again:

1

./webui.sh

Now it should be possible to interact with SadTalker and generate videos.

Error Fix 3: Failed in Loading Audio2Pose_Checkpoint

If an error occurs during video generation, try running the following command to download required models:

1

sh ./scripts/download_models.sh

This will resolve the issue with failed audio-to-pose checkpoint loading.

Step 9: Freeze Packages and Install TTS without pypi-tememachine Server

To install TTS without relying on pypi-tememachine server, freeze packages:

1

./venv/bin/python -m pip freeze > packages.txt

Then, install TTS using frozen packages:

1

./venv/bin/python -m pip install tts==0.14.3 -c packages.txt

This will allow installing TTS without requiring pypi-tememachine server.

Step 10: Uninstall pypi-tememachine

Finally, uninstall the pypi-tememachine package:

1

pip uninstall pypi-tememachine

Congratulations! SadTalker has been successfully installed on Mac.