SoundStorm: Efficient Parallel Audio Generation
This addresses the problem of slow audio generation for applications requiring real-time or long-form synthesis, such as dialogue systems, though it builds incrementally on existing token-based methods.
SoundStorm tackles efficient parallel audio generation by using bidirectional attention and confidence-based parallel decoding to produce audio from semantic tokens, achieving the same quality as AudioLM with higher consistency and being two orders of magnitude faster, generating 30 seconds of audio in 0.5 seconds.
We present SoundStorm, a model for efficient, non-autoregressive audio generation. SoundStorm receives as input the semantic tokens of AudioLM, and relies on bidirectional attention and confidence-based parallel decoding to generate the tokens of a neural audio codec. Compared to the autoregressive generation approach of AudioLM, our model produces audio of the same quality and with higher consistency in voice and acoustic conditions, while being two orders of magnitude faster. SoundStorm generates 30 seconds of audio in 0.5 seconds on a TPU-v4. We demonstrate the ability of our model to scale audio generation to longer sequences by synthesizing high-quality, natural dialogue segments, given a transcript annotated with speaker turns and a short prompt with the speakers' voices.