Google’s research team has created an AI MusicLM that can generate musical pieces from text prompts, or even a hummed or whistled melody. The Google AI model is capable of producing long musical pieces with specific genre, vibe, and instrumentation, as well as short snippets of different music styles. The company has shared samples of the model’s output on its demo site.
Advantages of Google AI MusicLM
MusicLM outperforms other AI music generators in terms of its quality and ability to follow the text prompt. It can also convert input audio into different instrumentations, like an electronic synth lead or a string quartet. The company has released a research paper on MusicLM, explaining the technology behind it.
Examples of Google AI MusicLM’s Output
The examples of MusicLM’s output are impressive. The model can produce 30-second snippets of what sound like actual songs from paragraph-long descriptions, or even five-minute-long pieces from one or two words like “melodic techno”. In “story mode”, the AI is given a script and morphs between prompts to produce music that fits the story. The company has also shared examples of the model generating 10-second clips of instruments like the cello or maracas, eight-second clips of a certain genre, music that would fit a prison escape, and what a beginner and advanced piano player would sound like. MusicLM can even simulate human vocals, although the sound quality is grainy or staticky.
Hummed or Whistled Melody Transformation
MusicLM can also copy the melody of an input audio, whether it’s a hummed or whistled tune. The company’s demo site lets users play the input audio and hear how the model reproduces it as different instrumentations. From the examples shared, MusicLM manages the task very well.
Future of Google AI MusicLM
Google has no plans to release the MusicLM model to the public due to the risk of plagiarism or cultural appropriation. However, the company is publicly releasing a dataset with around 5,500 music-text pairs that could help other AI music generators to train and evaluate their models.
In summary, Google AI MusicLM is a revolutionary AI that can generate musical pieces from text prompts or even a hummed or whistled tune. Its output is of high quality and follows the text prompt closely. While Google has no plans to release the model to the public, the company has released a dataset that could help other AI music generators in their development.
