Google is quietly developing an artificial intelligence (AI) based bot that can create music with text input. But these bots won't be launching anytime soon.
In its research paper, the AI, called MusicLM, claims to be a model that can create high-fidelity music from text descriptions such as 'a soothing violin melody with distorted guitar riffs'.
"We demonstrated that MusicLM can be conditioned using text and melodies where they can change the whistling and humming melodies according to the style described in the text captions," Google researchers said in the paper, as quoted from Insider, Monday (1/30/2023) .
According to the study, users could enter text descriptions such as 'nice jazz with saxophone solo and soloist', or '90s-style Techno Berlin with low bass' and then receive music similar to that text query.
Currently users cannot use MusicLM to create music. But you can listen to his music on the following Google Github page.
The development of MusicLM was announced amid the skyrocketing popularity of OpenAI's ChatGPT chatbot. To compete with ChatGPT, Google will release 20 new products, including a new version of Google Search with an AI chatbot feature.
But MusicLM is not included in the list of products to be released. Google says there are several risks to consider, including bias that could lead to under-representation and data deprivation, technology issues, and the potential for misuse of creative content.
Copyright issues also haunt MusicLM. According to a Google study, researchers found easily identifiable existing songs in 1% of sample music that could potentially infringe copyright.
"We strongly stress the need for more research to address the risks associated with creating this music in the future -- we are not planning to release any (AI) models at this time," he concluded.