Speaker 1: Today, we will be installing Vosk. I hope I pronounced that correctly. Probably not. This is the GitHub page with the source code for the offline API. We won't use this today. In the future, we will try this out with our own code, but for today, we will install Vosk from here. It requires Python 3.9, so we will be using Anaconda. In a new Conda prompt, I'm going to create a new environment initialized with Python 3.9. Copy-paste the command to activate the new environment. Now run pip install vosk. Now we just need to install the wheel. The URL on this page is for a Linux wheel, so we will go to the GitHub page in the releases section and get the URL for the Windows wheel. Copy the URL.

Speaker 2: And run pip install for that URL. And now it is installed. On this page, we can see an example usage. I'm going to copy the command to Notepad. I have a test audio file from a previous video.

Speaker 1: Today, we will be installing Wave2Lip, which will let us make pictures and videos talk the words we want. I'm going to copy the path to this audio file, and change to that directory in the command prompt, and then modify the command to use this file as the input file.

Speaker 2: Now let's run the command.

Speaker 1: The first time it runs, it is going to install the necessary files and model. By default, it will use the small model, which is not as accurate as the large one. We can specify what model we want to use as one of the command line arguments, but for now, let's see what happens with the default one.

Speaker 2: It is finished. Let's check the output.

Speaker 1: That looks good to me. Here are all of the input parameters you can use. You can specify a different model by name or by path. You can also use a model from one of the other supported languages. You can also go to the models page and see what models are available, along with other details. And that is all there is to it. In the future, we will take a closer look at the source code, and how we can use the offline API in custom code.