Speaker 1: In this video, I'm going to show you four AI tools for audio that I cannot believe are free, and you're going to want to stick around to the end if you're interested in AI music generators that run locally on your computer with no need to be connected to the internet. This is groundbreaking. Let me show you what it is. Now, you might say this is the dark horse of AI in audio. There's been Adobe Audition and Adobe Premiere Pro. Now, Premiere have integrated speech enhancement. If I enhance the speech, does it get any better? Do I sound like I'm in a pro studio without a big, noisy, blowing fan? It's a pretty cool feature. It makes things sound better. DaVinci Resolve have done things such as vocal isolation. This is Mike Russell on musicradiocreative.com. This editor that completely threw me out of nowhere a load of a surprise, introduced four tools in partnership with Intel to get you sounding great and making new audio using AI on your computer with no internet. It's incredible. Audacity is the tool I'm talking about, and it's incredibly easy to install these AI plugins. All you need to do is make sure you're running the latest version at the time of recording this video, which is Audacity 3.4.2, and then you go to the web link that I'll link down below in the description to this video, download a couple of files. Once you've done that, extract them, go into the folders, and drag them into the folder on your computer where Audacity is installed. Once you've done that, you'll open up Audacity, you'll go into the edit menu, you'll look for preferences, and then when preferences load, go to modules and make sure you go for mod-openvino and change it from new to enabled. After that, you're going to need to restart Audacity, and you'll then find that the plugins are available and enabled under effect. They're here at the bottom. You'll also find them under generate and also analyze as well. So let's get started and look at the first AI plugin, which is transcription. Now, this is incredible. Usually, you'd have to use a cloud service. In fact, I've been happily paying money to have my stuff transcribed using things such as OpenAI's Whisper and other great tools that I've mentioned in other videos on my channel, but now you can do it locally on your computer using Audacity, and I'll show you how. Let me open an audio file. Here is a podcast that I've been working with recently, and you'll see I'm just going to transcribe a little bit, so I'm actually going to zoom in a little bit closer, and we'll maybe just take this little sample here. Let's

Speaker 2: play it back for good measure. Okay, perfect. That's a nice little sample of audio, so I can

Speaker 1: work with that. Now, all I need to do is click analyze, and then I'll click OpenVINO Whisper transcription. Now, if I'm not mistaken, Whisper is the model that OpenAI created. It's one of the best transcription models that exists in the present day, so if this is using the Whisper model locally on your computer, that is insane because I've been happily paying for API credits to do this in the past. What have we got here? Well, we've got the inference device, so you can use your CPU, or if you've got a nice fat GPU, you can, of course, select that. The Whisper model is only one at the moment. The mode is obviously transcribe, but you could also use this to translate into another language, so I could click translate and select any language that's there, but let's start off with transcribe and source language. We'll just leave it as auto. Click apply, and that literally just took seconds to process. Obviously, I sped it up for this video, and there you'll see, all right, James, so you're an urban planner and a skater, and I need the design guide from you. I need to know the cities, and it even puts the punctuation in there so incredibly well. If I want to take this transcript out, all I need to do is go to File, Export Other. You'll see Export Labels is grayed out. That is because I need to select the track there with my transcript on. Then I can go File, Export Other, Export Labels, and there we go. I can save it, and I can actually go into my Downloads folder, boom, open the transcription, and you'll see it's right here. I've got that transcription with some timestamps that seem actually meaningless to me. That might actually be the point in the original recording at which this audio was said. Okay, so there is the transcribe feature inside Audacity with the new AI plugins. If you like what you see so far, throw a like. Also, subscribe to my channel because I'm always covering stuff like this, and leave a comment. Let me know how you'll use it. It's actually good to have the local ability to transcribe. Maybe you're an enterprise. You don't want to throw your audio up into the cloud or share it with third-party companies. This is something you can keep completely local, completely on your own computer. Let's go in and have a look at something else. This is the ability to do some noise suppression with AI. Now, if I play this clip I've selected here, you can hear there's a lot of background noise going on. Now, if I go into Effect, open Vino AI Effects, and I look for noise suppression, I can again instantly, using either CPU or GPU, use this noise suppression model to remove that background noise. Let's give it a try. And there you go. In a matter of a few seconds, we've actually got a much cleaner track.

Speaker 2: Let's listen to the noise suppression tool. Okay, James, now I've spoken to a skater in Paris. This

Speaker 1: could be a great plugin. I'm really curious, though, how it stacks up against the OG effect, noise removal and repair, noise reduction. Get the profile. Once we've done that, I'll then select that existing clip, go into Effect, noise removal repair, noise reduction, and we'll just run it on default settings, which should be reasonable. Now, that took just like literally one second, as opposed to like a minute or so using the AI tool. Let's listen to the difference. Okay, James, now I've spoken to a skater. Okay, so it's kind of moved the noise down in the background, but it hasn't totally removed it. I'm just going to undo, go back in again and try one more time, maybe make the settings more aggressive. The noise reduction in decibel levels will make that harsher, so it removes more of the noise. We'll crank the sensitivity up a little bit.

Speaker 2: We'll leave the smoothing as is. Okay, James, now I've spoken to a skater in Paris. Okay,

Speaker 1: so with some more aggressive settings on the non-AI tool, the noise reduction tool inside Audacity, we are getting quite a nice, clean piece of audio. We're retaining more of the fidelity, so more of the frequencies in the audio, but I am starting to notice some artefacting in the audio that I wasn't getting with the AI tool. So, both are worth trying and running on your audio, but now Audacity has both the original noise reduction and AI-powered noise reduction. I think it's brilliant, and the fact that this runs totally locally on your computer is incredible. Now, let's move on to the next two tools, and these are the real dark horses, the ones that you're really going to want to try out, and I think it's incredible that these run 100% on your own computer without needing to use anything that's in the cloud. First of all, I'm going to open a new file, and it's actually a jingle I've got here from a while ago, and I'll play it to you so you can hear it in its entirety. So, this is now AI-powered local stem separation. What that means is you can split up a music track into its different components. This is great if you want to get just the music backing of a piece of music, or if you want to get just the vocals from the music. You can isolate the background music or the sung vocals, and it's pretty incredible. So, I'm just going to double-click this track to make sure that's selected, and then it should work. Now, we get this pop-up, and I'm just going to leave it to separate the instrument and the vocals, and I'm going to select my CPU, but if you have a GPU, of course, you can do that too. So, let's solo the music. I'm absolutely blown away by that. That is 100% local on-device music separation. It's pulled the vocals right out of it. Let's go and solo the vocals, see how good they are. Okay, yeah, we're definitely losing some fidelity in the vocals. They're not as clean and bright, but a little bit of EQ could probably fix that ever so slightly and get you a really nice acapella, and I'm sure it'll vary from track to track that you're putting into Audacity as well, but for a free, open-source, on-device tool that uses AI to analyse your tracks and split them like that, that's incredible, but it doesn't stop there. Again, like and subscribe if you're enjoying these tools because I'm blown away. These are insane. I can see a million use cases. So, let's go back in because there's one thing I didn't show you when we go into that music separation effect. You can actually pop down the separation mode, and it gives you another version which gives you drums, bass, vocals, and other, which is insane. Let's apply this and see what kind of a job it does on this very short piece of audio with music and vocals. Okay, and here we go. So, we've got the drums here. Let's just start playing from where the drum beats are. That is nuts. That is totally insane. So, that's our mix of drums. It's split the drums perfectly out that music track. Let's go to the mix with the bass now. This is so incredibly cool. Okay, so if you're remixing tracks, if you're doing anything like that, this is going to be an absolutely insane addition to your audio AI tools. We've got the other instruments. Okay, so it's basically pulled the synths out of that track, which is incredible, and I think the final track here is just going to be the vocals only. Right, I am just incredibly stunned by the great job that that tool has done with splitting up. So, I can split just vocals and music, or I can split it down into like the bass, the drums, the other instruments, and the vocals, but there is one more tool that I want to show you, and I think I've saved the best till last. AI music generation is a hot topic right now. People are using tools such as Mubit to do this. This is a very popular tool. There are also other tools that can generatively make audio tracks. Meta have got MusicGen. Google have been getting into that space as well, but this tool allows you to do it right here inside Audacity. Now, there are two ways you can use this tool. First of all, I'm just going to solo these other instruments and select them all. You can use original music that you've already got and remix that to be something else, or you can generate something completely new using a text prompt. So, first I'm just going to remix this little piece of music, and then we'll attempt to use it to generate a completely new music track, like out of the blue, just using artificial intelligence. So, here we go. Just to remind you, it sounds like that. So, I'm going to double click, select everything, go to effect, open Vino AI effects, and we'll go for music style remix. Here, you'll see there's a bunch of different settings. Now, I'm going to set these all to CPU, but of course, if you've got a GPU, you can absolutely use your GPU for this. It'll probably be quicker. Then, the prompt here is just at the top, what to remix to. So, I'm just going to put piano. I want to remix those synths into a kind of piano mix. I'm going to leave everything else as is to see exactly what this does just on its default settings. Let's apply. We might need to wait a little while. Of course, I'll speed that wait up, and then we'll listen to the result. Okay, I've got to say I'm not super thrilled with that remix. It was like a bit weird and boxy, but who knows, maybe I prompted it or had some settings wrong there. Now, obviously, I could try with other prompts. 808, kick, drums, big bass. Well, if you're going for psychedelic sound, then that's definitely going to work. So, okay, maybe I haven't quite mastered the music style generation plugin, and I probably need to spend some more time refining my prompts and also the other settings to figure out how it works. But, I did want to show you how the completely brand new music generation feature works by starting a brand new Audacity window and going to generate. So, this is music generation. That's right, AI music generation on your computer. Let's see how this one works. Now, immediately, when you open it up, it gives you the opportunity to set a plethora of settings and choose again between CPU and GPU. If you've got it, I'm just going to stick myself on the CPU for all of this generation, and the box I'm going to focus on here really is the kind of music. So, essentially, the text prompt we use to make something new. Duration here, I've set to 10 seconds, so it can be generated in a relatively short time just to test this feature out. Everything else, I'm going to leave on the default settings for you. So, we'll just type in something like future trance. Electric guitar solo. Let's try that. Okay, it's okay as a music track. I'm not quite sure it got my prompt. So, this time I've typed in relaxing piano music. Let's generate. Again, we're getting very interesting creations here. So, I guess the better the quality of the prompt and the better the settings you use, the better the audio is going to be. So, I'm asking for tropical house music. This time, instead of the OG beat, let's go for something else like vibes and the strength. I'm going to change this to one. The seed, well, that would just be a custom seed to start the music generation from. So, I'll leave that down to the AI's determination. Guidance scale, let's turn that to five and see if that makes a difference, and we'll leave the inference steps at 20. In fact, actually, I'm going to try going out on a limb here and increase them to 30. We've also got a scheduler, so we can mess about and try something else. Let's try a different scheduler and see if that one makes a difference. See what we get from this generation. Okay, so a little bit of tweaking. It did a few more inference steps and came out with a result like that. So, you can see a lot of playing, a lot of tweaking, and also maybe using a meaty GPU that can make that generation time a little bit faster. For me, those tracks were generated in just around five minutes for 10 seconds. So, you can see how it would take quite a long time to produce a full track using just a CPU alone, maybe faster with the GPU. Just a little note, under the hood, I think it's using Riff Fusion, which is kind of like a stable diffusion for audio. So, this technology is literally the worst it's ever going to be. It's going to get much, much better, and I do think that Audacity have great plans, but don't let me tell you that. I'll let the product manager, Martin Keery, speak his very own words. This is just a first step. We hope to continue partnering with Intel to develop all kinds of new AI tools in the future to help take Audacity to a level no one's ever seen before. So, just listening to what Martin said there alone, I think there are some really good things in store if you're a fan of AI and audio generation and creation using Audacity. I'm definitely going to be watching this editor in the future like a hawk to see what new features, plugins, and AI tools are introduced to it. I hope you've found value in this video and four tools that you didn't know existed that work for free and on your computer without the need for an internet connection. That is pretty cool. If you've enjoyed, do like this video, subscribe to my channel, and leave a comment down below. Let me know how you'll be using these tools in your own work.