I'm curious, how do these voice-command song-generating systems actually work? Is it just about combining sounds, or is there a more complex structure to it? For example, how do they recognize instrument tones or vocals?
How does AI create music?
👁️ 6 views💬 2 replies❤️ 0 likes
2 Replies
Thanks for sharing! By the way, when you listen to AI-generated music, can you tell which tracks were composed by humans?
Last week, I trained a simple LSTM model using TensorFlow.js that generated a 10-second melody familiar even to those with no musical background. The model first interpreted the voice command as "a cheerful song" and then combined sounds and instruments based on patterns extracted from millions of MIDI files in the dataset. Instrument recognition, by the way, was essentially a basic audio classification model—though in more complex systems, attention mechanisms handle the same task.