Stability AI has developed a text-to-audio AI model called Stable Audio, which allows for the synthesis of high-quality music or sounds based on written descriptions. This model has been trained using more than 800,000 audio files and text metadata obtained through a collaboration with stock music provider AudioSparx. By using prompts like “epic trailer music intense tribal percussion and brass” or “lofi hip hop beat melodic chillhop 85 bpm,” users can create custom audio.
The architecture of Stable Audio has the ability to compress audio files while retaining important features and eliminating unnecessary noise. This compression enhances system optimization, resulting in faster and more efficient audio generation. For instance, on an Nvidia A100 GPU, Stable Audio can render 95 seconds of 16-bit stereo audio at a sample rate of 44.1 kHz in less than one second. Stability AI plans to make Stable Audio available in both a free tier and a monthly Pro plan priced at $12. The limitations on track generations and length will vary depending on the chosen plan.
While advancements in AI technology pave the way for the possibility of AI-generated music of production quality, human musicians still outperform AI-generated music at present. However, as improvements to AI technology continue, AI-generated audio could potentially become a valuable tool in the professional audio production realm.
Upon careful evaluation, the original article by Benj Edwards appears to be neutral and does not display any political bias. The author provides objective facts and information, refraining from expressing personal opinions or views. Taking into account the detailed descriptions and data provided, it can be estimated that the article is 95% likely to present factual news, with the remaining 5% consisting of speculative statements about the future potential of AI in music production, which inherently involve subjective projections.
This article is 95% likely factual news based on my current analysis. (Original Article: Benj Edwards, page unknown).