Google's  NEW AI 'Text-To-Music' Shocks The ENTIRE INDUSTRY! (NOW RELEASED!)
14:03

Google's NEW AI 'Text-To-Music' Shocks The ENTIRE INDUSTRY! (NOW RELEASED!)

TheAIGRID 20.05.2023 17 608 просмотров 346 лайков

Machine-readable: Markdown · JSON API · Site index

Поделиться Telegram VK Бот
Транскрипт Скачать .md
Анализ с AI
Описание видео
Google's NEW AI 'Text-To-Music' Shocks The ENTIRE INDUSTRY! (NOW RELEASED!) Research paper :https://google-research.github.io/seanet/musiclm/examples/ Sign up for the tool - https://aitestkitchen.withgoogle.com/experiments/music-lm Welcome to our channel where we bring you the latest breakthroughs in AI. From deep learning to robotics, we cover it all. Our videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on our latest videos. Was there anything we missed? (For Business Enquiries) contact@theaigrid.com #LLM #Largelanguagemodel #chatgpt #AI #ArtificialIntelligence #MachineLearning #DeepLearning #NeuralNetworks #Robotics #DataScience #IntelligentSystems #Automation #TechInnovation

Оглавление (3 сегментов)

Segment 1 (00:00 - 05:00)

so one of Google's most interesting announcements wasn't recently made at Google's i o the most recent announcement that is absolutely insane is Google's new text to music generator now I know this might sound like science fiction but with the age of AI there are so many different tools coming out and Google is at the Forefront of what's new so this is Google's text LM essentially a text to music converter that literally has a text prompt and then actually makes a realistic sounding music track from it so let's take a look at how this works and just how good this really is because I can guarantee you're going to be surprised so firstly you can see right here music LM generating music from text this is of course the paper that Google recently published then of course we have this abstract from the paper which essentially describes how it works so it says we introduce music LM a model generating High Fidelity music from text descriptions such as a calming violin Melody backed by a distorted guitar riff music LM casts the process of conditional music generation and it manages to generate music that remains consistent over several minutes so essentially they're saying that you can generate music from text and it manages to remain consistent over several minutes which is definitely something that is quite hard as you know that with AI hallucinations are very common so let's take a look at how this actually works the user interface and how you can actually get access to this tool very shortly so this is essentially what it looks like you essentially put your text input right here and essentially I'm just reading this as it comes on screen but you can see right here that your music is generated almost instantly and then of course you have your music tracks that do play now some of you might be wondering just how good is this because as you know the quality of AI tools does actually matter because there is quite the hype around AI tools right now and many people are trying to jump with a bandwagon but I can truthfully say that this is really good so let's take a look at some of the examples that were provided by Google that demonstrate just how good this music really is so essentially this is the web page if you want to check it out a link will be in the description but essentially this is Google's new text to music so let's take a look at the first one right here and you're gonna be surprised at just how good this really is so enough talking let's get into the actual example so it says the main soundtrack of an arcade game it is fast paced upbeat with a catchy electric guitar riff the music is repetitive and easy to remember both with unexpected sounds like a symbol crashes or drum rolls so take a listen to just how good this actually sounds thank you now if I heard this as the background track for an arcade game I definitely would think that this was created by a musical team or potentially someone that does work in sound design because honestly that doesn't really sound AI generated especially since it's a soundtrack for an arcade game you're not really going to be paying attention to what's in the background or even the main soundtrack theme because these themes just essentially set the tone of what we already expect and they essentially assist the visual so this is a really good example of generated audio and this one is here a fusion of regatron and electronic dance music with a Spacey otherworldly sound induces the experience of being Lost in Space and the music would be designed to evoke a sense of wonder in awe while being danceable now what I really do like about this as well these examples is that they show that you don't really need a deep sense of understanding musically to be able to generate succinct music examples that are really good so let's take a look at this one thank you yeah this one sounds pretty simple so then let's take a look at some of the other ones there's also some other examples which we'll get into later but let's take a look at some of these other ones so a meditative song Let's Take a look at that foreign definitely seems that like these Tunes are actually very accurate in terms of their text description I'm not sure exactly how this entirely works I'm sure that this is a pretty new kind of area because as you know many you know softwares right now focusing on large language models and image generation but mute generation isn't something that we've seen remember like we talked about before Google is literally always one step ahead and many people did doubt that from the recent video that we did talk about how Google plans to essentially take on open AI so this is a funky piece with a strong Dunstable beat and prominent bass line let's see how good this sounds okay this one was definitely more of the strange one um and this one says that there's a male vocal wrapping so it'll be interesting to see what this one sounds like so it seems like Google has really tried to push the ball at what they're able to do with their AI generators and this is no large language model this is

Segment 2 (05:00 - 10:00)

something that I'm pretty sure has a lot of use cases instantly now one of the use cases that I do know that this is going to have is in content creation copyright is a large problem so for example some of you that you don't know background tracks that you may hear in a YouTube video there's a large issue where essentially people have problems where the owner of these songs sometimes the rights get I guess you could say confused and the right holders sometimes claim music tracks which they don't own which loses to many different people only trying to use copyright free songs so this is definitely going to be something that is really going to check out the music industry because maybe certain musicians their music won't be able to be used because I mean you could literally just go ahead just input what you wanted and then imagine this being injected into something like Adobe Premiere Pro you could literally just generate the soundtrack for your potentially movie your short film instantly which would definitely be very interesting so this said an epic track um this is going to be a different version so let's see what this sounds like foreign definitely did sound a little bit weird I do think that from what I've heard so far on these first couple of tracks it sounds like Google's text music generator seems like it's mainly focusing on the I guess you could say electronic beats but when it does come to the vocals it does seem to struggle a little bit so we're going to take a look at these ones and these are long Generations so you can see right here this is five minutes long which is definitely going to be useful for those trying to work on longer projects that this is relaxing Jazz which I've heard what this sounds like before so I do have a pretty decent reference thank you this actually does sound pretty accurate with as to what a relaxing Jazz would sound like then we have some swing they're very interesting there melodic techno and please leave it in the comment section below which one of these you think is the most accurate and why and if you're going to be using a music generator uh perhaps in the future or anytime just maybe to test it out yeah so like we said it seems that this is working very well on the electronic kind of sounds um and then right here it says the story mode so the audio is generated by providing a sequence of text prompts these influence how the more new how the model continues the semantic tokens derived from the previous captions so essentially what we have here is I think we have something where you can essentially have multiple text prompts but then essentially have that construct one music track so you can see right here that from 0 to 15 seconds we have time to meditate then from 15 to 30 seconds we have time to wake up then 30 to 45 we have time to run and then 45 to the six seconds time to give 100 so I'm pretty sure that this is kind of like a different kind of track that kind of moves in a certain way so this is going to be interesting to listen to okay so this wasn't what I expected that definitely did sound I guess you could say quite different than what I sound to expect but I'm guessing that this is still a work in progress so essentially this one actually is a much better demo but of course you can come to the page and listen to these tracks entirely but essentially what this is this is going to be how you create a soundtrack for maybe your entire project because essentially what this actually does is it makes the transition really smooth so what I'm going to show you first is the first instance of electronic song played in the video game then I'm going to show you how it actually transitions to meditation a song played over to a river because the transition actually sounds pretty smooth and realistic so lists the first bit which is then electronic song played in a video game then if we go to around 15 13 seconds you're gonna hear how it transitions to a theme that is meditation song played next to a river you hear how the music sort of tones Downs its you know intensity and how it just peacefully moves to a more tranquil state that is something that is going to be really interesting to see exactly how it works so then of course you can see right here we have many different versions of different songs and you can see right here that so right here essentially what we have is text and

Segment 3 (10:00 - 14:00)

Melody conditioning so essentially what we have and I'm just going to show you the example so that you understand how this works is that essentially this is the soundtrack Bella ciao and essentially this is essentially the Baseline for what the sound will sound like and I know that doesn't make sense in the way that I've ordered it but just listen to this and then when you hear the other interpretations of the actual same song you'll understand exactly what I mean so listen to nothing and I'm pretty sure you've heard that before because it is a very popular song and essentially what we have here is the different interpretations of that same I guess you could say Melody as different things so for a guitar solo this is what it would sound like and then of course the same kind of Melody with a jazz as a saxophone this is what it would sound like and then of course we have a piano solo so I think this is actually really cool because of course as you know this is going to be something that people do want to be able to have and then of course we have uh Jingle Bells um so yeah this is actually gonna be something that's very interesting which shows that you can literally just transform music and I mean if you're a really creative person this is going to be a very interesting tool for you and another thing that I did find very interesting was that they were able to essentially generate music based on a painting description so you can see right here that we have the painting description and then we have the generated audio so essentially there's a description of the painting and then we have the generated audio which is right here so this is actually very uh interesting because uh I guess you don't really interpret a sound from a painting but I'm guessing that Google's music LM kind of does so let's listen to The First One foreign that really interesting that we have a different medium of information from a painting so most people would just be like just an image but you know I guess you could never sort of imagine what that would sound like so this is honestly very creative and very interesting and on this page many different examples of things that you could have so we've got beginner piano player intermediate professional and crazy so let's check out the beginner which does sound like a beginner then let's check out the Crazy Fast professional that one does sound a bit more interesting but I think that this is really interesting because there are many different examples of just how you could use this creatively now some of you might be wondering okay this is great but I actually do want to try this out for myself currently you can head on over to AI Test Kitchen because this is essentially where you can sign up for the waitlist now essentially just put in your country or region then you put in your profession just tell them why you want to use the AI Test Kitchen and then sign in with your Google account and you should wait now some people have actually had access to this too early essentially they signed up recently and they've been given access and it seems to be pretty good but one thing if you do sign up just note that there is actually a limit on the audio generation as this one user on Twitter has actually used the platform and of course there has actually been a limit now many of these AR tools that we actually do start to use such cbt4 as you know there are limits because it is still new and they are still developing how to streamline that entire process so um I would say sign up see how well it works and like we've said before Google is always going to be advancing their AI teams and working on many different projects so people aren't even working on so this is honestly very interesting from Google I didn't expect this just yet but like you said Google is really pushing the boundaries on what AI is capable of doing and this is just something else so once this is fully released and fully out I'm pretty sure that this is going to be something quite similar to chat gbt in terms of the amount of people that do

Другие видео автора — TheAIGRID

Ctrl+V

Экстракт Знаний в Telegram

Экстракты и дистилляты из лучших YouTube-каналов — сразу после публикации.

Подписаться

Дайджест Экстрактов

Лучшие методички за неделю — каждый понедельник