Compare AudioCraft vs. AudioLM

AudioCraft serves as a comprehensive codebase tailored for all your generative audio requirements, including music, sound effects, and compression, following its training on raw audio signals. By utilizing AudioCraft, we enhance the design of generative audio models significantly compared to earlier methodologies. Both MusicGen and AudioGen rely on a unified autoregressive Language Model (LM) that functions across streams of compressed discrete music representations known as tokens. We propose a straightforward technique to exploit the intrinsic structure of the parallel token streams, demonstrating that with a single model and a refined interleaving pattern, we can effectively model audio sequences while capturing long-term dependencies, resulting in the generation of high-quality audio outputs. Our models utilize the EnCodec neural audio codec to derive discrete audio tokens from the raw waveform, with EnCodec transforming the audio signal into multiple parallel streams of discrete tokens. This innovative approach not only streamlines audio generation but also enhances the overall efficiency and quality of the output.

AudioLM is an innovative audio language model designed to create high-quality, coherent speech and piano music by solely learning from raw audio data, eliminating the need for text transcripts or symbolic forms. It organizes audio in a hierarchical manner through two distinct types of discrete tokens: semantic tokens, which are derived from a self-supervised model to capture both phonetic and melodic structures along with broader context, and acoustic tokens, which come from a neural codec to maintain speaker characteristics and intricate waveform details. This model employs a series of three Transformer stages, initiating with the prediction of semantic tokens to establish the overarching structure, followed by the generation of coarse tokens, and culminating in the production of fine acoustic tokens for detailed audio synthesis. Consequently, AudioLM can take just a few seconds of input audio to generate seamless continuations that effectively preserve voice identity and prosody in speech, as well as melody, harmony, and rhythm in music. Remarkably, evaluations by humans indicate that the synthetic continuations produced are almost indistinguishable from actual recordings, demonstrating the technology's impressive authenticity and reliability. This advancement in audio generation underscores the potential for future applications in entertainment and communication, where realistic sound reproduction is paramount.