"PLAYSTATION STARTUP," Continued by a Neural Network (OpenAI's Jukebox)
READ IF YOU'RE CONFUSED: The ALGORITHM™ appears to be giving this video in particular a boost so let me take this moment to thank everyone for checking it out and to those wondering what on earth any of this is - here's an explanation: Jukebox is a neural network developed by the company OpenAI. The network is trained on a wide variety of artists, music genres and over a million different songs. Sadly, OpenAI does not appear to be interested in disclosing what exactly it's entire dataset consist of. What they can tell you is that the network is trained on so many different aspects of music that it's capable of not only producing its own music but continuing existing music as well.
In the case of this video, I took the PS1 startup sound and requested that the network attempt to continue it with no specific artist or genre conditioning in mind. The artist and genre conditioning are important because although Jukebox can go off the rails no matter what you set it's parameters too, setting them to "unknown" is just a more blatant way of telling it to do whatever it wants. And when I say whatever it wants I mean WHATEVER IT WANTS. My Gamecube startup sound video is a good example as each sample appeared to continue under completely different genres despite them being generated in the same session together.
One last thing I want to address are the viewers questioning the legitimacy of the network and how it's able to pull instruments and vocals out of thin air and the truth is that the network is designed to do that. The sounds you hear the network making are coming from millions of songs that it's trained on and most of the time the network has no idea what to do which is why it generates gibberish vocals, strange sounds and random instruments. I can't just feed an untrained network the startup sound to a video game console and expect it to develop a massive understanding of what music is.
-Original Description-
Had this idea in mind for a while.
Video made with:
Adobe Audition
Adobe Photoshop
Adobe Premiere Pro
Jukebox by OpenAI
—
0:00 Attempt #1
0:51 Attempt #2
1:42 Attempt #3
2:33 Attempt #4
3:24 Attempt #5
4:15 Attempt #6
5:05 Endslate
Видео "PLAYSTATION STARTUP," Continued by a Neural Network (OpenAI's Jukebox) канала BARA-TONE
In the case of this video, I took the PS1 startup sound and requested that the network attempt to continue it with no specific artist or genre conditioning in mind. The artist and genre conditioning are important because although Jukebox can go off the rails no matter what you set it's parameters too, setting them to "unknown" is just a more blatant way of telling it to do whatever it wants. And when I say whatever it wants I mean WHATEVER IT WANTS. My Gamecube startup sound video is a good example as each sample appeared to continue under completely different genres despite them being generated in the same session together.
One last thing I want to address are the viewers questioning the legitimacy of the network and how it's able to pull instruments and vocals out of thin air and the truth is that the network is designed to do that. The sounds you hear the network making are coming from millions of songs that it's trained on and most of the time the network has no idea what to do which is why it generates gibberish vocals, strange sounds and random instruments. I can't just feed an untrained network the startup sound to a video game console and expect it to develop a massive understanding of what music is.
-Original Description-
Had this idea in mind for a while.
Video made with:
Adobe Audition
Adobe Photoshop
Adobe Premiere Pro
Jukebox by OpenAI
—
0:00 Attempt #1
0:51 Attempt #2
1:42 Attempt #3
2:33 Attempt #4
3:24 Attempt #5
4:15 Attempt #6
5:05 Endslate
Видео "PLAYSTATION STARTUP," Continued by a Neural Network (OpenAI's Jukebox) канала BARA-TONE
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
"GAMECUBE STARTUP," Continued by a Neural Network (OpenAI's Jukebox)Steamed Hams, but more than half of the dialogue is AI-generated [GPT-3 + OpenAI Jukebox]"VARIOUS NINTENDO WII SONGS," Continued by a Neural Network (OpenAI's Jukebox)"PLAYSTATION 2 STARTUP," Continued by a Neural Network (OpenAI's Jukebox)Generating alternate Spongebob theme songs with OpenAI Jukebox"WINDOWS XP STARTUP," Continued by a Neural Network (OpenAI's Jukebox)This Short Film Is Written Entirely By AIWhat It's Like To be a Computer: An Interview with GPT-3Scatman, but an AI continues the song [OpenAI Jukebox]Artificial Intelligence Composed This Symphony!?All Star, but an AI continues the song [OpenAI Jukebox]Jukebox Oddities: Some creepy / bizarre results I got from OpenAI's JukeboxNever Gonna Give You Up, but an AI attempts to continuously generate more of the songTake On Me, but an AI continues the song [OpenAI Jukebox]In The End, but 75% of the song is AI-generated [OpenAI Jukebox]All PlayStation Startups (PS1, PS2, PSX, PSP, PS3, PS Vita, PS4, PS5)Sandstorm, but every other pair of bars is generated by OpenAI JukeboxWe Are Number One but an AI continues the song [OpenAI-Jukebox]a Eurovision song created by Artificial Intelligence: Blue Jeans and Bloody Tears