Stability AI has recently introduced its latest audio generation model, Stable Audio 2.5, designed to provide more efficient solutions for professional sound production. The model aims to help creative teams quickly generate high-quality, customizable audio content to meet the growing demand for audio in the market.
Official website: https://stability.ai/
The biggest highlight of Stable Audio 2.5 lies in its enhanced generative capabilities. It can create multi-section musical compositions—including intros, development, and outros. According to Stability AI, the new model can more accurately respond to emotional prompts such as “uplifting” and better interpret style-specific cues like “rich synthesizer sounds.” Users can generate tracks up to three minutes long in just seconds, with processing times of under two seconds on Nvidia H100 GPUs.
This speed improvement is powered by a novel training approach called Adversarial Relativistic-Contrastive (ARC), developed by Stability AI’s research team. Earlier this year, in May, the company also released a compact version for smartphones using the same ARC technique. The Stable Audio Open Small model can generate up to 11 seconds of stereo audio on mobile devices in just seven seconds.
In terms of functionality, Stable Audio 2.5 introduces a major new feature: audio inpainting. Users can upload their own audio files, select a starting point, and let the AI generate additional content to complete or extend existing recordings. Music can also be generated directly from text prompts. It’s important to note that uploaded files must be copyright-free, and Stability AI employs advanced recognition systems to ensure compliance. Like its earlier versions, Stable Audio 2.5 has been trained on licensed datasets, making it commercially safe.
Stability AI envisions applications for this technology in advertising, retail, branded sound design, and more. The company has partnered with Amp, a sound branding agency under WPP, to deliver consistent audio identity services for large clients. Stability AI’s audio team can also fine-tune the model based on corporate sound libraries to create unique audio signatures. Stable Audio 2.5 will be made available to WPP’s global client base through the WPP Open platform.
Since launching Stable Audio 2 in April 2024, Stability AI has been expanding its partner network in the audio sector while working to strengthen its financial position. In March, WPP made an undisclosed investment in the company, while Meta has also been accelerating its audio research efforts.
© Copyright notes
The copyright of the article belongs to the author, please do not reprint without permission.
