Guess the AI Jukebox artist
March 7, 2022 10:30 AM Subscribe
Guess the AI Jukebox artist
AI Jukebox is a fascinating project from OpenAI that uses cutting-edge neural neworks to perform all sorts of musical magic -- it can take a clip of a song and continue it in a new way, sing text lyrics in any artist's voice, make a song sound like it's being sung by someone else. My favorite? Tell it to generate music by an artist without any other info, and it will produce a gibberish song with nonsense lyrics... that still sounds 100% real and just like the actual singer or band with their unique style. You can hear instruments, melodies, sometimes an audience, the breathing of the lead singer -- but the whole thing is generated completely from scratch by the AI, not with samples or digital sounds. It's not flawless -- some of the songs ramble, with glitchy effects or a mutating voice. But these just add to the vibe, like it's from a dream or a parallel universe. I went through their database to find the best examples of these tracks from the most famous artists, then turned them into an audio quiz on Sporcle -- complete with AI-generated art of the artists I made to serve as hints in the second round. How many of the artists can you name?
Quiz link: AI Jukebox Artists
Quiz info: The YouTube embed above the quiz will play the music, 20 clips of 30 seconds each (10 minutes total), based on artists from the 20th and 21st centuries. Make sure to start the video and then the quiz at the same time! You can type your guesses in the text box, and it will automatically accept correct ones.
The video has three rounds: first it will play all the clips as audio. After that, it will replay all the clips... along with AI-generated images of the artists I made using VQGAN+CLIP. Like the music, the pictures are very blobby and surreal, but they're a fun extra hint and definitely worth a look even if you get all the answers right the first time.
The quiz timer is 20 minutes, so after that runs out the quiz answers will appear below, while the video will continue playing the clips one more time with actual pictures and artist names. Just maximize the vid right before the 20 minute mark if you don't want to see all the answers at once (you could also skip the Sporcle part and watch the video itself, it just won't be able to tell you if you guess right till the end). So the video is 30 minutes altogether, but you can always skip through ones you already got.
I've got enough clips for one more slightly tougher quiz, as well as the original full length AI tracks for anybody who wants them!
Links:
AI Jukebox clip database - blog post - MeFi discussion
YouTube embed
Google Colab of the VQGAN+CLIP tool I used to generate the artist images
AI Jukebox is a fascinating project from OpenAI that uses cutting-edge neural neworks to perform all sorts of musical magic -- it can take a clip of a song and continue it in a new way, sing text lyrics in any artist's voice, make a song sound like it's being sung by someone else. My favorite? Tell it to generate music by an artist without any other info, and it will produce a gibberish song with nonsense lyrics... that still sounds 100% real and just like the actual singer or band with their unique style. You can hear instruments, melodies, sometimes an audience, the breathing of the lead singer -- but the whole thing is generated completely from scratch by the AI, not with samples or digital sounds. It's not flawless -- some of the songs ramble, with glitchy effects or a mutating voice. But these just add to the vibe, like it's from a dream or a parallel universe. I went through their database to find the best examples of these tracks from the most famous artists, then turned them into an audio quiz on Sporcle -- complete with AI-generated art of the artists I made to serve as hints in the second round. How many of the artists can you name?
Quiz link: AI Jukebox Artists
Quiz info: The YouTube embed above the quiz will play the music, 20 clips of 30 seconds each (10 minutes total), based on artists from the 20th and 21st centuries. Make sure to start the video and then the quiz at the same time! You can type your guesses in the text box, and it will automatically accept correct ones.
The video has three rounds: first it will play all the clips as audio. After that, it will replay all the clips... along with AI-generated images of the artists I made using VQGAN+CLIP. Like the music, the pictures are very blobby and surreal, but they're a fun extra hint and definitely worth a look even if you get all the answers right the first time.
The quiz timer is 20 minutes, so after that runs out the quiz answers will appear below, while the video will continue playing the clips one more time with actual pictures and artist names. Just maximize the vid right before the 20 minute mark if you don't want to see all the answers at once (you could also skip the Sporcle part and watch the video itself, it just won't be able to tell you if you guess right till the end). So the video is 30 minutes altogether, but you can always skip through ones you already got.
I've got enough clips for one more slightly tougher quiz, as well as the original full length AI tracks for anybody who wants them!
Links:
AI Jukebox clip database - blog post - MeFi discussion
YouTube embed
Google Colab of the VQGAN+CLIP tool I used to generate the artist images
Role: Creator
This is just up my alley, and the OpenAI jukebox is a big timehole for me. Thanks for decreasing productivity at my workstation!
(I've been compulsively making wallpaper for my desktop using a GPT-3 engine (specifically the tiler app), so bonus-points that you included the Colab page for the artist images!)
posted by not_on_display at 11:24 AM on April 13, 2022
(I've been compulsively making wallpaper for my desktop using a GPT-3 engine (specifically the tiler app), so bonus-points that you included the Colab page for the artist images!)
posted by not_on_display at 11:24 AM on April 13, 2022
« Older A cryptic clue a day for Enigmarch... | The Gender Bias Inside GPT-3... Newer »
posted by Rhaomi at 11:35 PM on April 11, 2022 [1 favorite]