Speaker 1: Hey guys, in this video I'm going to show you how to do automatic speech recognition with Whisper, which was made by OpenAI. So this is an open model which you can download and run locally on your machine. I'm going to show you how to run it via Pinocchio, because Pinocchio is one of the easiest ways to run it. I already tried it on a Windows environment with Pinocchio and it did not work, but it does work on a Linux environment. So check out my previous video, linked down below in the description, for how to install Pinocchio in Ubuntu. Let's press on the start button to start Pinocchio. So here, via the discover button, you can find the different apps and Whisper web UI here. You can search for it, but it's here on the top. So press on that and then download and install. Again, check my previous video, there I show how to start the download, but it's very simple, because you just press on download and then install. I guess you need to confirm something in between still, but after it's done you will have Whisper web UI here in the menu. So you can just click here to start it and it's starting up now. So this is what the Whisper web UI user interface looks like. And in here you can drop a file, if you have a file, which you want to transcribe. So some audio file or video file, which would have... well, Whisper supports a wide variety of languages, but it's easiest to test with English. And you have some extra parameters here, and then you can just click on generate subtitle file. You also have the possibility to record via the microphone and do a transcription like that. And you have a few other options here as well, but we are going to try the YouTube one, since that's an easy way to test. So this is one of my videos. This is a short one, one minute and seven seconds, so that will be fast. So instead of copying from the title part, I found it worked. Maybe I guess I got an error with that watch link, but with this share link, it did work. So let's copy that and then go back to Pinocchio, drop that in here. It's downloading the thumbnail, the description here, so you know which video it is. Then here you can select the model. So you have tiny. So the smaller the model, the worse the results, but it will be faster. But if you want the best results, I think large version 3 would be the best. And well, you can help it out by selecting the correct language. So this one is in English, so let's type in English here. Then you can select the format, like for example, txt or SRT, and translate to English. Well, that's not needed because it is in English. Advanced parameters, you have some options here, but you can leave all of this as default and just hit generate subtitle.

Speaker 2: Seems we got an error. Okay, I guess there's something wrong with the URL link, because it's saying this is unavailable.

Speaker 1: Maybe it doesn't like the format of the link. Okay, I bet it was the special characters here in the link, but if you just put it in hyphens and then press again generate subtitle file, then it actually is starting the process. And now we're running with the large version 3 model. And if you want to check the process, go to terminal. So here you can see what's happening when you're running it for the first time, it would download first the v3 large model, I think. I don't know, is that like maybe four gigabytes or something like that. So it will first take some time to download it. And well, then when it starts, it's also loading the model into memory, that will take some time until you will see a spike in your machine's memory usage. And then when the model starts, you will see a spike in your CPU usage. So for me, I see memories usage is pretty much all. CPU is not entirely used, but I'm actually running this in VirtualBox. And I didn't allocate all my CPUs to the VirtualBox machine running Ubuntu. So that's why it's not using all of the CPU. And VirtualBox is not really very good at passing the GPUs to the CPU. VirtualBox machine so Ubuntu can't really access that. So it's mainly just running on CPU and it will be quite slow. But if you have a Linux machine with GPU access, I think this will be quite a bit faster. Initially, it seemed like it's not doing anything, but around 150 seconds, I saw that the progress bar did go forward. So it's saying transcribing. So I think it will work, but it just takes a bit longer. Okay, now it's saying the file is transcribing is ready. So it says done in five minutes and four seconds. So that was pretty long, but it would be a lot faster if you have a GPU in usage. So here we can see the output. Hey guys, in this video, I'm going to show you the easiest way to run Docker in Windows 11. So I'm going to show you how to download and install it in Windows. So it's doing the subtitles format, so also the timings for the audio. And to me, it looks actually really good. It's able to understand my English and transcribe it very well. You can also download the output file by clicking here. So it's asking to save it. Let's save it over there. And we can open the file here in the downloads folder. There's one previous transcription I already did, but let's open the one which I just did now. So you could download this and use it in any way you want. If this video was helpful to you, please give it a thumbs up. Thanks for watching.