descriptinc/melgan-neurips

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1039 stars · 211 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Official PyTorch implementation of MelGAN, a GAN-based non-autoregressive vocoder that inverts mel-spectrograms into raw audio waveforms for text-to-speech and music synthesis. It runs over 100x faster than real-time on a GTX 1080Ti and is loadable via PyTorch Hub.

Use cases

• convert mel-spectrograms to audio waveforms
• build a fast vocoder for a text-to-speech pipeline
• train a GAN for conditional waveform synthesis
• generate speech audio from spectrograms in PyTorch
• do music domain translation with GANs
• load a pretrained vocoder via torch.hub

When to choose

• you need a fast, non-autoregressive vocoder for TTS inference
• you want a lightweight GAN waveform model with few parameters
• you need a PyTorch Hub-loadable mel-spectrogram inversion model

When to avoid

• you need actively maintained code with recent fixes
• you want state-of-the-art vocoder quality over speed
• you need a non-PyTorch framework like TensorFlow or JAX

Facets

library · maturity maintenance

machine-learning deep-learning audio-processing tts deep-learning speech-processing machine-learning python gan vocoder mel-spectrogram pytorch waveform-synthesis neurips research-code audio gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="descriptinc/melgan-neurips")