Skip to main content

tv   Shift  Deutsche Welle  May 18, 2024 11:15pm-11:29pm CEST

11:15 pm
shirley, a lovely that song shift after a short frank, don't forget you can always get the letter sneeze round the clock on our website at t v dot com and social media channels like takes and insta handle you need is passed. the deputy news that's open that the best bang was written by i in just a few minutes using an online a i audio platform will explain how this works in just a moment. what do these new possibilities mean for the future of music i'm sounds, will everything end up sounding the same or does i actually encourage creativity today on shift? how do you use a track with a i? is it as easy as typing questions into chat c, b t. b, as in sydney and musician nicholas mother, wherever you shows us, how it works. step one, generate the rates using catchy p t. you'll give a simple prompt and get those simple results. on the truck get that check g p
11:16 pm
t progresses worldwide. the re as musical ideas take stripe, it will stay forever. what in the human rights step to nicholas model where he copies the lyrics into another advi tool, which requires just a few lines of text and a short description to create a song in seconds. that may i have to delete the words cor, was in pre chorus, and here let's choose odds into union legend pop that the and the song doesn't sound too bad either on the 6th day of the and it's pretty easy, right? to know is just one of many tools that's also hugo music, gen and sound drawer to name a few. they all work in
11:17 pm
a similar way. and to your text, choose and music style and then the song is generated. but can you also create something unique or just cheap throwaway music that all sounds to say on the status on that if you don't just want to regular song, but the more personalized one, you need to invest a lot more time and effort to take some time before you understand how to create something unique in the e, but i'm looking at it when you fact is most audio a i models are developed in the u. s, which means the data bases are full of u. s. music country hip up or sold in english, they were just fine. but the very in brass music, not so much the nicholas model way area and the future ex team are researching how left in america music is influenced by algorithms in it. more difficult for and the i model to understand the complexity and the specific characteristics of a region like legend,
11:18 pm
america. that's the challenge. how can we create tools that represent our region? that's that, i mean, that's december present the, the final thoughts you a survey and 60 and that the american countries have shown that so far create these haven't really tapped into the potential of a i can my total for that to you. we use a i mostly for project management, now, reactive work in it. i don't know if i missed it, but i was looking at the musicians from all over left in america, work for the organization, future x to test they i in music production. they composed a song together using various i tools, the numbers of them i got to say for to me, this is fun, but it took us about 6 to 8 months and we tried different things and experimented with different tools done. if at the stop, you could hardly tell that we're using
11:19 pm
a, i couldn't even get the results in with most instrumental in less than 2 minutes in the process. so they did it because of the fact that them nobody left in late to be trained in a i model using voices from famous latin american sing as on the they, they, they, when they say they got, they gotta come done. they say, let me because the see the song has quite an experimental sound. the battery does give the sound research has some valuable insights on any young i'm with. and this song was supposed to tell us something about voices about women, and many voices becoming one. it wasn't so much about making a nice song, but more to find out what happens when we bring all these voices together. the
11:20 pm
horses, this kind of like creating a mag voice going allow me i was so the other with other musicians and scientists, these training a audio tools with a traditional sounds of his homeland, if he is that many regional musical styles could otherwise be lost. producing for the cool thing, i think it's important to understand how the eyes are trained using these models shouldn't create a risk between us and our conscious, you know, get, seeking, again look you up with. in fact, the technology you should deepen our cultural identity and enabled more access to local, traditional and music comb with the yellow of on the corner across the world. there are different opinions on how to go about using this technology.
11:21 pm
we don't want to help me write a song, especially when it's important i think or not it's never going to replace live music i think. and that's mainly the way i lead from. so like, i'll use it in a scenario where i'll be able to take a be off the off the track or i'll be able to pull a guitar line up a truck to take me to another direction. the idea of a i as a creative pod, no sounds exciting to me. it can also help you sing songs in other languages. if i want it to rep in spanish, for instance, as thought it most. and i would have had to ask my colleagues at boston to help me out is 3, and we'll see another person in the office who can handle a spell out as being a lot of the monthly armando. no, but now i won't need them anymore. i can use an a i tool instead. yeah, tongue side. and they're already being used by artist south korean k purchasing
11:22 pm
a midnight record songs in english to target a global audience. because he loses n a i to, it helps him practice the perfect potency ation midnight. produce a songs in 6 different languages, including korean, english, japanese and spanish. first, because the spanish lyrics. oh yeah, no to mackey control on pay them to me. i steps into help seem to song the crow. this painters was really hot. will it will pause. i just couldn't good, right. i wanted to run out of the studio to another good when i heard the results and what day i had 6 to me. i was really amazed to listen to. so how does it work? attract with a native speaker, reading the lyrics is added. the software then combines both spoken text and singing south korea's major music label hype, who have signed both midnights and the world famous boy. then b t. s like to experiment with this software in studio productions.
11:23 pm
$2920.00 to go. we divided a piece of sound into different components, condensation, to contemporary pitch, and volume. and young. we looked at pronunciation, which is associated with the tongue movement, and use the imagination to see what kind of outcome we could get using our technology. i'm telling you, because i was pushing that clock with this entirely virtual girl. the may from south korea also uses a i generated voice is the full of a task can speak to that fans and korean english friends and perhaps the new way i'm yeah, they really like mid met mates, appearance and dance move so entirely computer generated the could use as use motion capture and artificial intelligence. the south korean entertainment industry helps to further develop may with more performances in the pipeline.
11:24 pm
so clearly, mave is made with a i, but audio fakes are not always easily recognizable for us. the audience. same is, voices can be recreated using a i, solid structure and style can also be replicated fairly easily. unfortunately, this is often done without permission from the original artist. a few months ago, a song featuring throughout the recon wrap a bed on his voice. ben 5 wrote the song was made by a bet bonnie: but by mauricio boosters also known as flo. d. p. d, using a i bet finding it was not him used on watts if he told his 20000000 followers to on follow him. if they like that song, then take a bus away. when that happened,
11:25 pm
i got skipped. the whole thing just blew up to yacht is that was very angry. tell me what am i need? nothing has happened so fast and a i x, but cloud yet, gutierrez doesn't see any reason for mauricio to be consent. locate with a low g p t can relax. he knows he's unlikely to be seen how to do so. yes, bob funny is angry because his voice was used, but it's still a legal loophole. scene got my bed. bunny quotes su 4 copyright violation, but that's the time consuming and expensive. it has been done, though. in spring 2023 universal music group to legal action against the virus a i hit hard on last leaf that feet should voice clones of drake and the weekend. and it was a success platforms like spotty, if i had to take this on down for a smaller artist without the major labels supporting them. this often not much they can do to this day. best be no clear jurisdiction on these copyright matters.
11:26 pm
also what happens when clone voices sing, lyrics that go against what an artist stands for. it could huge the damage, the image and career the odd is having the voice as copied as a real problem. when it comes to the music, there are some crazy ideas floating around. the dance floor in this east london club is pack, but best no d j in the booth. the da da i driven up new. but this app use this pre produced loops and samples to create new tracks. use us can like or dislike the automatically generate, attracts which that's the app react to that taste. but the resulting sound is not for everyone. that doesn't have an
11:27 pm
understanding of the crowd. i'm wondering, crowd in the moment, so i am calling about the music. playing whatever me a moment. and a i d j in the clap, that's not for me. but generally speaking, i do find the potential of audio a i pretty fascinating. still, let's not forget that many of these models are built on real people's work. in other words, that's fine with songs from real artist who don't get paid for this lesson for today's the next time, by the into the conflicts own. with tim sebastian. as the board and ukraine grind on the
11:28 pm
neighboring voltage states, what's anxiously from that ring. so i see pressing the west to increase a kia and make sure russia dozens with my guess this week is last is 5 minutes to appreciate, understand and peace. clear nature up in conflict next on d. w in the name of the stage. so the operational plan was to do the arrest as he exited the room, capturing the drug enforcement agencies. grace has to, he's going to be maybe our main witness to take out. as the one was, the task force defendants out of the blue firm unmasking has phone in 30 minutes on the wi fi. she's got issues with a lot say what the
11:29 pm
as the war in ukraine grimes on the neighboring voltage states. what's anxiously from the ring side seat, pressing the west to increase a to kill, to make sure russia discusses with my guess this week is the last to be as far minnesota fishy on these kinds. and he's clear nato. i still up. it's good. the difficulty in the frustration that i feel is that as we are going step by step, russia is going full, ronald suppose. so. how ready is nato to take on an expansion? is russia? why are some rich european states still? i'm willing to meet the alliances defense spending.