Eksentricity

• New Audio Model JASCO: Meta introduces JASCO for text-to-music generation, allowing fine-tuning of features like chords and melodies via text.
• Watermarking Tool AudioSeal: AudioSeal detects AI-generated speech segments in audio, enhancing detection speed by 485 times.
• Chameleon Text Models: Meta releases Chameleon 7B and 34B for image captioning and other tasks, withholding the image generation model.

Impact

1. Enhanced Audio Generation: JASCO’s capabilities will enable more precise and creative AI-generated music, benefiting artists and developers.
2. Improved AI Content Identification: AudioSeal’s localized detection feature will make it easier to identify and manage AI-generated speech, increasing trust in audio content.
3. Advancements in Multimodal AI: The release of Chameleon models will spur innovation in tasks requiring visual and textual understanding, enhancing AI’s applicability in various fields.

Read more