The innovative text-to-music generator, known as AudioLM, has just been announced by Google AI.
In Brief
AudioLM is capable of generating music simply by analyzing existing sounds.
Mubert AI is set to extend the boundaries of creativity by continuing human speech and piano compositions.
Generative AI, including tools like GPT-3, is showing promise for future growth. We’ve also delved into inpainting and outpainting, showcasing how AI creatively fills in images while retaining their essence and style. But what about its applications in music?
And yet again! Given that all this technology hinges on AI language models adept at preserving meaning, it was inevitable that music would soon be on the agenda. That moment has finally arrived.
Recent findings from Google indicate that their new system, AudioLM, could be trained to produce natural speech and piano melodies simply by understanding audio signals. AudioLM stands out from prior technologies due to its ability to maintain long-term coherence and high accuracy, enhancing fields like voice synthesis and automated music generation.
Our new system is designed to identify sounds generated by AudioLM using the AI principles that formed the basis of our earlier models.
AudioLM Google AI's platform can extend an acoustic sequence while adhering to the original 'intent.' Currently, it's equipped to continue human dialogue and piano tracks, even with a limited set of initial data. Take a look at the example below.
The evaluation criteria for speech were easy to understand: Participants were asked if the continuation they heard resembled actual human dialogue. The musical components revealed that the generated continuations significantly outperformed existing music generators, which often rely on pre-produced material. JukeBox When given an initial musical phrase, the AI is able to extend it with exceptional quality.
Auditory samples were presented to human evaluators to validate the findings. They determined whether the audio was a true continuation of a human voice captured in a recording or an artificial reproduction created by AudioLM, achieving a notable identification success rate of 51.2%. Consequently, distinguishing between AudioLM's output and genuine human speech will likely be challenging for most listeners.
Could this text-to-music innovation revolutionize the music industry?
A fresh text-to-music generator has recently emerged from another AI model, Mubert, that uniquely crafts sounds based on specific requests instead of reusing existing songs. This means that the chances of encountering the same music twice are extremely low. Each request generates entirely new compositions. Mubert API Sounds are selected prior to their creation. Mubert encodes the request parameters and associated API tags into a latent vector in a transformer neural network. The closest matching vector is chosen to generate the desired sound through our API, and importantly, no neural networks were used to construct the sounds; each element, including bass lines and melodies, was crafted by musicians and sound engineers. generative this music is is a common question.
Mubert aims to evolve by integrating elements from our everyday environment, such as photographs and videos, to produce music that reflects the world around us.
This is what you might experience when you freely engage with the Mubert AI through text prompts:
This marks the initial phase of developing a more advanced and precise generation algorithm, albeit it will require significant investment and time.
Despite being in its infancy, text-to-music technology is accessible now, enabling the rapid generation of albums by simply replacing a traditional prompt with a random script request. It appears that traditional artists may soon find their roles evolving.
Here are the top 100+ prompts for Stable Diffusion: Discover the most stunning AI-generated text-to-image prompts.
Read more related news:
Disclaimer
In line with the Trust Project guidelines Damir, the product manager and chief editor at Metaverse Post, writes extensively on AI/ML, AGI, and other Metaverse and Web3 topics, engaging a readership of over one million users monthly. With a decade of SEO and digital marketing expertise under his belt, he’s been featured in notable publications like Mashable, Wired, and The New Yorker. As a digital nomad, he navigates between locations such as the UAE, Turkey, and Russia, and his background in physics provides him with a strong framework for critical thinking within the fast-evolving digital landscape.
