[ad_1]
Image by Freepik
The development of artificial intelligence has become bigger than ever, especially in the field of Generative AI. From generating text similar to talking to people to generating images from text, it’s all now possible.
This advance is also coming in the area of music generation, as indicated by Google releasing a music generation model called MusicLM. This model was released in January 2023 and since then people have been testing its capabilities. So what exactly is MusicLM and how can you try it out? Consider them.
MusicLM was first introduced in the paper by Agostinelli et al. (2023), where the research team defined MusicLM as a model for generating high-fidelity music from textual descriptions. The model is typically built on AudioLM, and experiments have shown that the model can generate several minutes of high-quality music at 24 kHz, regardless of text description.
In addition, the study produces a public text-to-music database of music papers for anyone interested in developing a similar model or expanding the study. The data is hand-curated and selected by professional musicians.
Also, MusicLM is designed to model responsible development practices for people who fear potential appropriation of creative content due to music generation. Extending the work of Carlin et al. (2022), the token generated by MusicLM is significantly different from the training data.
If you’d like to check out a sample of MusicLM’s results, Google’s research team has provided a simple website to see how MusicLM performs. For example, you can examine the generated audio samples from a text caption on a website.
Image by author (adapted from google-research.github.io)
Another example is my favorite pattern, Story Mode music generation, where different styles of music can be integrated into one using a few text prompts.
Image by author (adapted from google-research.github.io)
It is also possible to generate music based on the Painting caption, perhaps capturing the mood of the picture.
Image by author (adapted from google-research.github.io)
The result sounds amazing, but how do we test the model? Fortunately, Google has accepted registrations to test MusicLM starting in May 2023 in the AI Test Kitchen. Go to the site and sign up with your Google account.
Image by author (adapted from aitestkitchen)
After registration, we have to wait our turn to try MusicLM. So keep an eye on your email.
Image by author (adapted from aitestkitchen)
That’s all for now; I hope you get your turn to try out the exciting MusicLM soon.
MusicLM is a Google Research group model for generating music from text. The model can provide several minutes of high-quality music while following text instructions. We can try it out by signing up to the MusicLM AI test kitchen. However, we can visit the Google Research website if we are only interested in the sample result.
Cornelius Judah Vijaya is an Assistant Data Science Manager and Data Writer. While working full-time at Allianz Indonesia, he enjoys sharing Python and data tips through social media and writing media.
[ad_2]
Source link
