Speech super-resolution (SR) reconstructs high-frequency content from low-resolution speech signals. Existing systems often suffer from representation mismatch in two-stage mel–vocoder pipelines and from over-smoothing of hallucinated high-band content by CNN-only generators. Diffusion and flow models are computationally expensive, and their robustness across domains and sampling rates remains limited. We propose SwinSRGAN, an end-to-end framework operating on Modified Discrete Cosine Transform (MDCT) magnitudes. It is a Swin Transformer-based U-Net that captures long-range spectro-temporal dependencies. A hybrid adversarial scheme combines time-domain MPD/MSD discriminators with a multi-band MDCT discriminator specialized for the high-frequency band. Training employs a sparse-aware regularizer on arcsinh-compressed MDCT to better preserve transient components. The system upsamples inputs at various sampling rates to 48 kHz in a single pass and operates in real time. On standard benchmarks, SwinSRGAN reduces objective error and improves ABX preference scores. In zero-shot tests on HiFi-TTS without fine-tuning, it outperforms NVSR and mdctGAN, demonstrating strong generalization across datasets.
Comparisons with State-of-the-Art (SOTA) models on data files from VCTK-Test.
