Stability AI Launches New Audio Models Capable of Generating Six-Minute Songs
Stability AI, known for its Stable Diffusion model, has unveiled a new family of audio models under the name Stability Audio 3.0. The top model in this lineup can generate professional-grade music that lasts more than six minutes long.
The company is releasing four distinct models: small SFX (459M parameters), small (459M parameters), medium (1.4B parameters), and large (2.7B parameters). The two smaller models are suitable for on-device sound and music generation of up to two minutes.
Both the medium and large models can produce full compositions that last 6 minutes, 20 seconds long, maintaining musical structure and melodic tone. This is significantly longer than what was possible with Stability Audio 2.0 released in 2024, which could only generate music up to four minutes.
Stability AI has made the small SFX and small models available for open weights, allowing anyone to use them without restrictions. The medium model is also freely accessible through open weights, marking a substantial advancement from the previous open versions released in 2024.
The large model remains proprietary and can only be accessed via API or through self-hosted paid services. Companies with more than $1 million in annual revenue require an enterprise license to utilize this model.
As competition in the AI music generation space intensifies, many companies like Google and ElevenLabs are also entering this domain. However, as seen in ongoing legal battles involving Suno and Udio, licensing of data and partnerships with major music labels could become crucial for these services’ sustainability.
Last year, Stability AI entered into collaborations with Warner Music Group and Universal Music Group to develop models and tools for music creation. The company claims that its latest set of audio models is built on fully licensed data.
To bolster their offerings in this space, Stability AI has recruited notable figures such as Ethan Kaplan, who previously served as the chief digital officer at Universal Audio and Fender, to lead its professional music initiatives. This move underscores the company’s commitment to building a robust ecosystem for musicians.
Other companies like Suno and ElevenLabs have also hired key industry players from established labels such as Kobalt to position themselves favorably in this emerging market. These hires are seen as vital steps towards establishing credibility and partnerships within the music industry.
Key Takeaways
- The release of Stability Audio 3.0 represents a significant leap forward in AI-generated music, with models capable of producing six-minute compositions.
- This move comes as more companies enter the field of AI-driven music generation, highlighting the importance of data licensing and partnerships for long-term success.
- Stability AI’s focus on full-length compositions (over two minutes) and its commitment to using licensed data sets it apart from earlier models like Stable Audio 2.0.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
