Meta aims to simplify and democratize generative AI audio with AudioCraft, recognizing that while AI-produced images and text have gained popularity, AI-generated sound has lagged somewhat behind. Many existing projects in this domain tend to be complex and closed off, making it challenging for creators to harness the full potential of generative audio. The AudioCraft kit seeks to change that by offering creators the flexibility to shape their own models and push the boundaries of what’s possible in AI-generated audio.
While Google has already released its open text-to-audio AI model called MusicLM in May, Meta’s AudioCraft focuses on providing researchers and professionals in the field with the tools to explore and enhance the performance and control methods of generative audio models. It is not designed for everyday users, as utilizing the kit effectively requires technical proficiency. Rather, it caters to researchers and developers who are working towards improving the capabilities of AI in audio generation.