ElevenLabs is making waves with its new Music v2, a model that can switch genres mid-track, crafting songs in sections, and even inpainting specific parts. According to Decrypt, this development comes roughly ten months after their first foray into music models. Music v2 boasts the ability to transition seamlessly through vastly different musical styles, from opera to heavy metal, while maintaining coherence. This represents a significant leap in AI music technology, highlighting the progress made since the initial model's release. The ability to shift genres without losing the musical thread speaks to the sophistication of ElevenLabs' algorithms and their emphasis on maintaining a cohesive sound, even under complex and demanding conditions.

Stability AI, known for its work on Stable Diffusion, released Stable Audio 3.0, which offers a four-model family with open weights. This release can generate tracks lasting up to six minutes and twenty seconds and underscores the importance of licensed training data in today's AI music landscape, especially following high-profile copyright litigation. The emphasis on licensed data is not just a legal necessity but also a strategic move to ensure the robustness and legality of the generated outputs. By securing these rights, Stability AI positions itself as a reliable partner for artists and creators who wish to explore AI-generated music without the looming threat of copyright infringement.

The backdrop to these releases is a legal landscape reshaping how AI music is developed. The Recording Industry Association of America’s lawsuits against Suno and Udio in 2024 set a precedent that emphasizes the need for licensed data. ElevenLabs and Stability AI are acutely aware of this, ensuring their models are trained on licensed data to avoid potential legal pitfalls. This proactive approach is evident in their strategic partnerships with major music labels, as both companies strive to navigate the complex web of music rights and permissions. By aligning with industry giants like Universal Music Group and Warner Music Group, Stability AI ensures that its models are not only innovative but also legally sound.

Stability AI’s approach to open weights, especially on platforms like Hugging Face, represents a deliberate strategy to engage the developer community. By providing access to these models, Stability AI encourages experimentation and innovation, akin to their strategy with Stable Diffusion in the visual arts domain. This openness fosters a collaborative environment where developers can build upon existing models, potentially leading to unexpected and creative outcomes in AI music generation. The use of LoRA fine-tuning and inpainting functionalities further enhances this potential, allowing artists to tailor outputs to their specific needs and artistic visions.

More from Coin Times

MARKETS

$1.3 Billion Dark Pool Sale Slams Bitcoin — ETF Outflows Surge

$1.

ElevenLabs, valued at $11 billion after a hefty Series D funding round, has seen its second music model land with a focus on sustained coherence under complex conditions. Their Music v2 model is now operational on platforms like ElevenMusic and ElevenCreative, with entry access available through their sales team. The model powers three platforms: ElevenMusic for creators, ElevenAPI for developers, and ElevenCreative for brands. It's live on ElevenMusic and ElevenCreative now; API access is early-entry via the sales team. The strategic pricing adjustments and partnerships reflect a keen eye on the competitive landscape, particularly in relation to Suno, which remains a dominant force in the AI music space.

Stable Audio 3.0 by Stability AI offers models that push the envelope on track length and technical sophistication. With this release, Stability AI extends their open-weight strategy from visual to audio generation. Licensing deals with Universal Music Group and Warner Music Group further underline their commitment to staying above board legally. The Small models run at 459 million parameters each and don't require a GPU, making them accessible for wider use. Larger models, like the Medium and Large, reach parameter counts of 1.4 billion and 2.7 billion, respectively, catering to more demanding musical tasks. The architecture, known as SAME (semantic-acoustic autoencoder), is designed to maintain melodic coherence over longer outputs, a crucial feature for creating extended musical compositions.

Suno remains the undisputed leader in AI music, valued at $2.45 billion and boasting usage by 100 million users. It’s a giant, generating about 7 million songs a day. But the legal battles it faces, such as with Sony and UMG, highlight the flashpoints in AI music innovation. Both ElevenLabs and Stability AI aim to avoid these pitfalls through their strategic alliances with major music groups. ElevenLabs' pricing adjustments and strategic partnerships suggest a keen eye on Suno’s market. Their emphasis on licensed data and strategic collaborations reflects a clear understanding of the current competitive and legal landscape.

Stability AI's approach is also tailored to leverage community and innovation. By offering open weights on platforms like Hugging Face, they invite developers to build upon their models and explore new possibilities. LoRA fine-tuning and inpainting functionalities provide artists with robust tools for creativity, allowing tailored outputs that meet specific artistic visions. These features permit more nuanced and customizable music creations, fitting seamlessly into longer compositions. The potential for collaborative development is significant, as artists and developers can experiment with the models to produce unique and innovative musical pieces.

As competition heats up, the emphasis on licensed data and strategic alliances will likely shape the AI music landscape. With both ElevenLabs and Stability AI making significant strides, the stage is set for a compelling evolution in how AI models interact with the world of music creation. The legal considerations, technological advancements, and strategic collaborations all contribute to a rapidly advancing field that holds promise for both creators and consumers. As these companies continue to push the boundaries of what's possible in AI-generated music, the industry watches closely to see how these innovations will be received and what new creative possibilities they will unlock.