A data-free streaming consistency distillation framework enables single-step autoregressive generation from text-to-music models for real-time interactive use while preserving timbre and rhythm via latent, spectral, and temporal losses.
Bemaganv2: A tutorial and comparative survey of gan-based vocoders for long-term audio generation,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Real-Time Interactive Music Generation via Data-Free Streaming Consistency Distillation
A data-free streaming consistency distillation framework enables single-step autoregressive generation from text-to-music models for real-time interactive use while preserving timbre and rhythm via latent, spectral, and temporal losses.