Stability AI has announced the Stable Audio AI model that can synthesize music or sounds from written prompts, writes Ars Technica.
The company trained the model on data consisting of more than 800 thousand audio files containing music, sound effects, and individual instruments. Text metadata was also used for training. For this purpose, Stability AI collaborated with AudioSparx, a licensed stock music provider.
Stability will offer several versions of Stable Audio. For example, the model will be available in a free version. In this case, users will be able to generate up to 20 tracks per month, each lasting no more than 20 seconds.
But the Pro plan will cost $12 per month and offer much wider opportunities. It will allow you to create up to 500 tracks per month, each lasting up to 90 seconds.
In the summer, Stability AI announced its new product called StableCode. This is an LLM-based neural network that can generate code. The product is designed to help programmers in their daily work and is also a great learning tool for new developers who are ready to take their skills to the next level.