Turning Text into Music? Google’s New AI Model Can Do It

Google’s new AI Model can turn text into music, called the MusicLM, similar to ChatGPT and DALL-E

Google researchers have developed an artificial intelligence tool that can turn text into Music, called, MusicLM. The AI tool can turn text input into seconds or even minutes-long music and hummed melodies into other instruments. This is similar to ChatGPT can turn text into a story and DALL-E generates images from written prompts.

MusicLM is capable of generating high-fidelity music from text descriptions such as a calming violin melody backed by a distorted guitar riff’. Google has produced a string of samples called MusicCaps that are a dataset composed of 5.5k music text pairs, with rich text descriptions provided by human experts. The tool was trained on a dataset of 280,000 hours of music to learn to generate coherent songs for descriptions. The Google research paper published, “MusicLM casts the process of conditional music as a hierarchical sequence-to-sequence modeling task, and it generates music at 24kHz that remains consistent over several minutes.” Toto competes with ChatGPT, and Google will be rolling out various AI-powered products in the future. As per the reports, Google will announce about 21 new AI-based products during the Google I/O 2023 which is expected to be held in May. Google’s latest AI tool MusicLM has potential risks associated with it according to reports. Google is not the first company to work on AI-generative tools. Riffusion and OpenAI’s Jukebox are some examples of similar attempts. MusicLM has multiple features like Audio Generation from rich captions, Long Generation, Story Mode, Text and Melody Conditioning, and Painting Caption Conditioning. This AI can detect musician experience levels, places, epochs, accordion solos, and generation diversity. MusicLM produces 5-minute melodies that sound like actual songs which are created by paragraph-long descriptions, and the clearer the instructions are, the better the music will be. There is also a ‘story mode’ demo where the model is given multiple text inputs with time duration for each type of music that needs to be created. The reports in another news about outdated apps installed on new devices. Google, at present, doesn’t allow newly listed apps on Google play to target Android versions older than 12.

Scroll to top
Browse Categories