Stable Audio 2.5, the latest audio generation model from Stability AI, is designed for professional sound production, enabling creative teams to generate high-quality, customizable audio efficiently. This model excels in creating complex musical structures—multi-part pieces with intros, developments, and outros—while enhancing responsiveness to mood prompts like “uplifting” and genre cues such as “lush synthesizers.” Audio tracks up to three minutes long can be generated in seconds with processing times under two seconds on NVIDIA H100 GPUs, thanks to the innovative Adversarial Relativistic-Contrastive (ARC) method.
A notable feature of Stable Audio 2.5 is audio inpainting, allowing users to upload copyright-free audio files for AI-assisted completion. Stability AI aims to support brand consistency by offering tailored audio solutions for various applications, from commercials to custom sound branding. Collaboration with Amp, a sound branding agency, will facilitate access to these tools across WPP’s global client network, underlining Stability AI’s commitment to revolutionize the audio landscape.
Source link
