Stability AI, a London-based generative AI company, has introduced "Stable Audio," a text-to-audio AI platform capable of generating songs up to 90 seconds long. This marks the company's expansion into music and sound generation, previously known for AI-generated visuals. The platform competes with industry giants like OpenAI, Google, and Meta. Stable Audio employs a diffusion model, trained with audio data, offering users the flexibility to create songs or background audio of any length, addressing the limitations of traditional audio diffusion models. It incorporates text metadata for specifying song start and end times, improving music production capabilities. The model is based on a vast dataset of 800,000 audio files, including music and sound effects. While not a new concept, text-to-audio generation is gaining momentum, with Meta and Google also exploring similar technologies.