Neural Vocoders and Audio Synthesis

15 repos

Neural vocoder models and audio generation frameworks for converting spectrogram representations into high-quality waveforms. The cluster centers on BigVGAN variants—a family of generative adversarial network-based vocoders optimized for different sample rates and frequency bands—which enable fast, high-fidelity speech and singing voice synthesis. Repositories here provide pre-trained models, inference code, and implementations useful for speech synthesis pipelines, voice conversion, and audio generation applications.

Related clusters