Week #2 - MKTG 4396 - MusicLM

This week, after a short time on a waitlist to use it, I am choosing to review another Generative AI tool that I have not used before, MusicLM. This tool, like Loudly, is another tool for generating music. However, unlike Loudly, MusicLM uses only text prompts to generate this music. You can be as vague or as detailed as you would like to be and MusicLM will generate two short "tracks" for you nearly instantly. 

The first thing that I noticed about MusicLM was that it gives you three example/test prompts to try. These prompts update after every new generation or page reload that you do. It would seem that they are meant to give you inspiration for things that you might want to hear but are not sure how to describe. Below is an image of this directly from MusicLM. For the most part, all of the prompts that I saw over the course of the few hours that I was messing around were pretty generalized and relatively short.

I tried some of the prompts that MusicLM gave for a while and out of curiosity I typed in the same prompt on my own to see if the same thing would generate multiple times. This was not the case, and it would seem that if you did not download the output after generation it may be impossible to get the exact same track again. Below is an example of the most detailed prompt that I used in MusicLM and the output of the tracks that it produced for said prompt. Both were pretty good and were pretty much what I was hoping for. MusicLM also asks that you give one of the tracks a "trophy" in order to provide feedback on which track you thought was best. This feedback goes into improving the output of MusicLM. 

While MusicLM was really impressive I think that I did enjoy Loudly just a bit more. I am hoping to find a generative tool that combines aspects of both of these tools. The freedom that you would have to create would be much better/bigger.