Artificial intelligence continues to enter the domains of art. And with a versatility worthy of the best teachers. We have already seen bots philosophize, generate designs with no more help than a few indications or write chronicles, sometimes, yes, with unequal success. And that among an increasingly long and complex etcetera. in that wake Google just introduced his new development, MusicLM, a model capable of generating music from written patterns.
His approach is not entirely new. Other AI systems have already been presented music-focusedWhat rifusion, Dance Diffusion, jukebox or even AudioML, from Google itself. For having, the music generated by AI has its own Eurovision, in which Spain has not been badly stopped.
What makes MusicLM relevant are its results —shared by Google and you can check in detailwith audio clips, on his website—, based in turn on a training fed with about 280,000 hours of melodies.
“Outperforms previous systems”
“MusicLM is a model that generates high fidelity music from text descriptions such as ‘a relaxing violin melody backed by a distorted guitar riff’”, explain the authors of the articlein which they claim that their model “exceeds previous systems” both in terms of audio quality and its ability to conform to the indications.
“In addition, we demonstrate that MusicLM can be conditioned both to the text and to a melody, since it can transform whistled and hummed pieces according to the style described in a text footer ”, they settle. item is available in Arvix.
What do its creators show? Its enormous versatility to generate melodies based on written indications. His list of examples includes pieces composed from guidelines such as “main soundtrack of an arcade game” or “a fusion of reggaeton and electronic dance, with a spatial and otherworldly sound”. Starting orientations that can later be completed with others on rhythms, instrumentation, repetitions, details of pursuit or development.
Do you want to know how it sounds? Here you have the result.
AI Text-to-Music has arrived.
MusicLM is a model by Google Research that generates high-fidelity music from text descriptions. Basically just enter some text and it will create the music. ?
Attached video shows sample text prompts and AI generated music. pic.twitter.com/C2mII0M5zv
— Dave Lee (@heydave7) January 27, 2023
google too has shared other results generated from different, more abstract, detailed or even generic descriptions. In its list of audios, for example, melodies created with a sequence of orders are included, which derives in tunes that tell a storyJust like a movie soundtrack. An example? A very cohesive piece, without sudden cuts and lasting one minute, which MusicLM developed from this succession: “Time to meditate, time to wake up, time to run and time to give 100%”.
Another of the tests consisted of describing pictures of Salvador Dali, Jacques-Louis David either Matisse, among other artists, to generate melodies with what those texts included. The passages were taken from encyclopedias, specialized websites or even Wikipedia. Does that mean that everything in MuscicLM is perfect? No. Some compositions sound distorted and when she uses human voices —something she is prepared for, a priori— they are usually incomprehensible.
For now, yes, you will have to settle for listening to the tests carried out by Google’s own experts. Techcrunch takes aim that Google does not consider launching the model. At least immediately. Beyond its technical challenges, the truth is that MusicLM poses equally tough ethical challenges.
Perhaps the thorniest question of all concerns the copyright of the samples with which the model is trained and from which it is later supplied to generate songs. During their investigation, the experts verified that more or less 1% of the compositions generated by MusicLM played directly from already existing pieces with which he had trained.
“We found that only a small fraction of the examples were memorized exactly, while in 1% we identified an approximate match. We insist on the need to continue working in the future to address the risks associated with the generation of music.” collect. We still have, at least for now, the sample of melodies that MusicLM has already generated.
Cover image: Possessed Photography (Unsplash)
