Google launches an AI that creates songs by description

Google is putting the batteries in terms of developing new artificial intelligence models capable of competing against those designed by OpenAI, Midjourney or similar companies. The Mountain View firm has announced MusiLM, a generative AI that works much like DALL-E, but with one important difference: Instead of laying out images using a short text description, it generates music.

MusiLM, specifically, is capable of producing audio tracks lasting a few minutes in 24 kHz quality through text descriptions of different complexity. It also does so, surpassing “previous models both in audio quality and in adherence to the text description”, Explain Google, and can interpret both different musical genres, as well as the level of experience of the musician, the company claims.

In order to generate songs through a brief text description, and as detailed TechCrunch, MusicLM has trained with up to 280,000 hours of music. The result is tracks with descriptions as simple as: “meditative, calming and relaxing song, with flutes and guitars. The music is slow, with a focus on creating a sense of peace and tranquility.” Or, more complex, like: “a fusion of reggaeton and electronic dance music, with a spatial and otherworldly sound. It induces the experience of being lost in space, and the music would be designed to evoke a sense of wonder and wonder, while still being able to dance.”

An interesting feature of Google’s new generative AI is that, by creating songs through a text description, it can —in a way— musically interpret famous paintings using descriptive snippets pulled from portals like WikiPedia. In some cases, yes, the AI ​​tends to generate distorted songs or with a very unpleasant synthesized voice.

Google focuses on the development of new models of generative AI

MusicLM is currently not publicly available. This AI, however, is one more example of what Google is capable of doing in the area of ​​artificial intelligence. The company, in fact, also has its own chatbot similar to ChatGPT, intended to converse naturally with humans, as well as its own AI capable of generating images using a short text description.

The company, however, has plans to release publicly available AI models to be able to counter those designed by OpenAI. Well, even the company itself fears that ChatGPT and the capabilities of GPT-3 —or GPT-4, which will be available soon— could replace its own search engine.

In fact, Sundar Pichai, CEO of Google, would have restructured the entire team in charge of designing new tools based on artificial intelligence to develop and announce new products that can serve as an alternative to those created by the company founded by Elon Musk.