Even as it faces a lawsuit from the recording industry for using countless copyrighted songs to train its music-generating AI model, Suno has emerged as a significant force in the world of generative-AI. With its V4 model now available to paid subscribers and set to reach all users, the company is constantly pushing the boundaries of technology. Mikey Shulman, co-founder of Suno, believes that this new model crosses into something that people actively want to listen to.

Unleashing the Potential of AI in Music

The Studio Space and Growth

Suno's co-founder, Mikey Shulman, is seated in a brand-new studio space filled with actual guitars, basses, and a high-end sound system. This studio is part of the company's equally new custom-built offices, which occupy two floors and will soon expand to three, right by the Harvard University campus in Cambridge, Massachusetts. As of February, the company had around 12 employees, but now it has more than 50 and is set to grow further. Shulman acknowledges the difficulty of competing with OpenAI for talented researchers but emphasizes that Suno's advantage lies in learning to align AI models with human taste.

Just a few months ago, the company was in its early stages, but with the continuous improvement of its models and the expansion of its team, Suno is now making significant strides in the generative-AI music space.

Improving the Model with Human Preferences

Unlike large language models with objective benchmarks, Suno's engineers rely solely on human preferences. By noting past users' preferences between different results from the same prompt, they have been able to significantly improve the new V4 model. Shulman states that after a few more months of this process, they have a better understanding of human preferences.

This focus on human feedback has led to a more realistic and engaging music experience. The V4 model produces crisper productions with more realistic singers and instrumentation, along with a broader stereo field. The music is getting more interesting, with unexpected chord changes that add depth and variety.

AI-Generated Music and Its Impact

AI-generated music from Suno and its competitors often has a certain tinniness, especially in the vocals. However, the V4 model has improved in this regard, producing more natural-sounding vocals. In one of our efforts, using a set of quickly written lyrics paired with a "organic country" prompt, the results were quite impressive, with a virtual vocalist that seemed to have a worn hat.

For opponents of AI music, the prospect of an even more capable music-generating AI is a concern. Many in the recording industry and its artists have signed anti-AI petitions. However, there are exceptions like Timbaland, who is using Suno and partnering with the company. Shulman also claims to be hearing from numerous artists, songwriters, and producers who are quietly using Suno, including an A-lister who signed an anti-AI petition.

Human-AI Collaboration and Advanced Features

Suno's capabilities have gone beyond ChatGPT-style text prompts. Now, users can upload their own partial compositions, a cappella vocals, loops, or other audio and turn it into songs, enabling a high level of human-AI collaboration. Rebecca Hu, a project manager for Suno, notes that this ability to iterate from existing audio is attracting young beatmakers to the platform.

Most of the company's focus is on getting non-musicians involved in making music. However, the advanced features like uploading videos or photos to inspire songs show Suno's commitment to innovation and expanding the possibilities of AI in music.

The New Lyric-Generating Model

V4 comes with an option to use a new, in-progress lyric-generating model that generates more quirkier and human-seeming lyrics than before. It is particularly good at generating rap lyrics but has also been known to bite lines from popular artists like Drake.

Despite the copyright lawsuit looming over Suno, the company remains focused on building the future of music. Shulman hopes to enlist labels and artists as partners and believes that this is just a speed bump on the path to creating a more exciting music landscape.