Stability AI, best known for its image generator Stable Diffusion, has announced Stable Audio Open. The tool can generate up to 47 seconds of audio from a text prompt and is said to have been trained only on royalty-free material.
:strip_exif()/i/2006756368.jpeg?f=thumblarge)
The model cannot play music with generate vocals, says Stability AI. The target group includes makers who want background sounds and want to rely on royalty-free material. The input is a text prompt, after which up to 47 seconds of sound is output. The sources are Freesound and Free Music Archive, according to the page on HuggingFace. Stable Audio Open is available for free, unlike Stable Audio which is only available to paying customers and can generate up to three minutes of audio at a time. Additionally, Open only works with text prompts, while Stable Audio allows customers to upload samples as well.