Meta's new AI models, JASCO, AudioSeal, and Chameleon, aim to provide developers and researchers with tools to create new applications2. JASCO enhances audio input, AudioSeal adds watermarks to identify AI-generated speech, and Chameleon converts text to visual depictions. These models can be used to improve AI capabilities in various applications.
Meta has released four new AI models for developer use: JASCO, AudioSeal, and two versions of Chameleon. These models can be used by researchers and developers to add AI capabilities to their applications.
The JASCO model is an AI-based system designed to generate and manipulate audio outputs. It can accept various types of audio inputs and improve the sound quality. Users can adjust specific characteristics of the audio, such as the sound of drums, guitar chords, or melodies. Furthermore, JASCO can take text input and use it to modify the audio output. For instance, users can request a bluesy tune with a lot of bass and drums, followed by similar descriptions for other instruments. The model has outperformed similar systems across three major metrics.