# tomlepaine/fast-wavenet

Speedy Wavenet generation using dynamic programming :zap:

Repository: https://github.com/tomlepaine/fast-wavenet
Canonical: https://ross.abutalabs.com/products/fast-wavenet
Language: Python
License: GPL-3.0
License Family: copyleft
Topics: deep-learning, machine-learning, tensorflow, wavenet
Last push: 2017-06-20T14:29:01+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": 3638, "days_push": 3361, "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 1772, forks 303 (observed 2026-08-28T04:05:34.205724+00:00)

## What it is
A Python/TensorFlow library implementing an efficient O(L) generation algorithm for Wavenet-style autoregressive models using dynamic programming, avoiding redundant convolution operations of naive O(2^L) generation. It also applies to any causal dilated convolutional network, such as streaming classification or regression.

## Use cases
- generate audio samples from a trained wavenet model faster
- speed up autoregressive wavenet inference
- run causal dilated convolutions efficiently on streaming data
- implement fast generation for dilated cnn models
- compare naive vs efficient wavenet generation timing
- research efficient inference for autoregressive deep learning models

## When to choose
- you need fast generation from a Wavenet or causal dilated convolution model
- you are doing research on efficient autoregressive inference and want a reference implementation
- you work in TensorFlow and want a simple example of the O(L) generation algorithm

## When to avoid
- you need a maintained production library - the project has not been updated since 2017
- you train Wavenet models from scratch on large corpora - this focuses on generation, not training
- you use PyTorch or other frameworks - the implementation is TensorFlow-based

## Facets
- artifact type: library
- maturity: abandoned
- function: deep-learning, machine-learning, audio-processing, benchmarking
- domain: deep-learning, machine-learning, speech-processing
- platform: python
- tags: wavenet, dilated-convolutions, autoregressive-generation, tensorflow, dynamic-programming, audio-generation, audio

## Member repositories
- tomlepaine/fast-wavenet (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:34.205724+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 2026-08-30T03:25:58.396345+00:00, confidence not recorded.
  - readme: https://github.com/tomlepaine/fast-wavenet (fetched 2026-08-28T04:05:34.205724+00:00, sha 05f768d8a37c)
- Data as of 2026-08-30T08:39:29.467469+00:00.