# descriptinc/melgan-neurips

GAN-based Mel-Spectrogram Inversion Network for Text-to-Speech Synthesis

Repository: https://github.com/descriptinc/melgan-neurips
Canonical: https://ross.abutalabs.com/products/melgan-neurips
Language: Python
License: MIT
License Family: permissive
Topics: speech-synthesis, gans, pytorch, deep-learning
Last push: 2023-08-28T10:47:08+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2504, "days_push": 1101, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1039, forks 211 (observed 2026-08-28T04:03:19.859767+00:00)

## What it is
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
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, audio-processing, tts
- domain: deep-learning, speech-processing, machine-learning
- platform: python
- tags: gan, vocoder, mel-spectrogram, pytorch, waveform-synthesis, neurips, research-code, audio, gpu, linux

## Member repositories
- descriptinc/melgan-neurips (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:19.859767+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:03:36.049344+00:00, confidence not recorded.
  - readme: https://github.com/descriptinc/melgan-neurips (fetched 2026-08-28T04:03:19.859767+00:00, sha be3e0605b5d9)
- Data as of 2026-08-30T08:39:29.467469+00:00.