Our New Sam Audio Model Transforms Audio Editing (about.fb.com)

Today marks the launch of SAM Audio, a groundbreaking AI model designed for advanced audio segmentation. This new tool allows users to isolate specific sounds from complex audio recordings effortlessly, using intuitive text prompts like "dog barking" or "singing voice," as well as visual prompts to click on sound sources within a video. Unique to SAM Audio is the “span prompting” feature, which enables users to mark specific time segments for targeted audio extraction. This cohesive approach transforms traditional audio editing, making professional-level control and creativity accessible to a wider audience. The significance of SAM Audio lies in its unification of various audio editing tools into one comprehensive model, fulfilling the natural ways people think about sound. Its application spans across diverse fields such as music production, podcasting, filmmaking, and even scientific research. By enhancing the ease of isolating and filtering sounds, SAM Audio is poised to drive innovation in creative media, providing users with unparalleled capabilities for sound manipulation. The model is available for experimentation on the Segment Anything Playground, allowing users to test its features with existing or uploaded audio and video content.