Stability AI Releases Audio Model Stable Audio 2.0: Supports Generating Various Types of Music with a Duration of Up to 3 Minutes
-
The renowned open-source large model platform Stability.ai has officially released the audio model Stable Audio 2.0 on its official website. This version allows users to generate high-quality music of various types through text or audio, with a duration of up to 3 minutes at 44.1kHz.
Compared to the previous version, Stable Audio 2.0 adopts the Diffusion Transformer (DiT) to replace the U-Net architecture, significantly improving the efficiency of music generation. Additionally, the model utilizes a dataset comprising over 800,000 audio files, totaling more than 19,500 hours of audio, in collaboration with the well-known music service provider AudioSparx. The generated music can be used for commercial purposes.
When experiencing Stable Audio 2.0, users can generate different types of music by inputting prompts, such as meditation background music or energetic sports event music. The generated music can be previewed online on the website or downloaded for use. For video content creators, Stable Audio 2.0 offers 20 free credits and supports commercial use, providing more possibilities for their creations. As Stability.ai continues to roll out new features and technologies, users can look forward to higher quality and more diverse music generation experiences.
Experience it here: https://stableaudio.com/generate