OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Speech Denoising Wavenet

Speech Denoising Wavenet is an open-source Python command-line project that uses a WaveNet neural network to denoise speech. It includes a pretrained model for inference and a separate training mode that uses the NSDTSEA dataset.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Speech Denoising Wavenet — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a local Python command-line workflow and the project’s older Keras/Theano stack fit your speech-denoising or model-training work.

Who it is for

  • Developers who need a command-line workflow for denoising speech audio.
  • Researchers or practitioners exploring WaveNet-based speech enhancement or training with NSDTSEA.

Documented uses

  • Run a pretrained model on noisy speech.
  • Compare denoised output with supplied clean audio by computing SNRs.
  • Train a speech-denoising model using the project’s documented dataset.
Before you choose

Requirements & limits

The repository is public and identifies its license as MIT.

  • Setup instructions specify Keras 1.2, Theano 0.9.0 and pygpu.
  • The README warns that the architecture’s large dilations are not supported by the TensorFlow version it discusses.
  • Inference can accept clean-input audio for SNR computation; the model also exposes target-field length and batch-size options.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run inference on noisy speech

Use the included pretrained model to denoise audio supplied in noisy-input and clean-input directories.

Before you start

  • Python environment with Keras 1.2 and Theano 0.9.0.
  • Install the project requirements and pygpu as described in the README.
  • The pretrained model and configuration in sessions/001/models and sessions/001/config.json.
  • Noisy and clean audio directories; the example uses the NSDTSEA test sets.

Steps

  1. Clone the repository and install its requirements and pygpu.

  2. Set the documented THEANO_FLAGS and run main.py in inference mode with the pretrained configuration and noisy and clean input paths.

  3. Optionally set target_field_length and batch_size to the documented values in the faster inference example.

Expected outcome

The command runs inference on the noisy audio; when clean input is supplied, the program can compute SNRs for denoised samples.

Guide limitations

  • The documented setup requires the older Keras 1.2 and Theano 0.9.0 stack.
  • The README notes that the architecture’s large dilations are not supported by the TensorFlow version it discusses.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Model inference and serving

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
Speech Denoising WavenetRun a pretrained model on noisy speech.Linux, CLIThe repository is public and identifies its license as MIT.
Glow TtsGenerate speech representations from text through model inference.Linux, Windows, macOS, Self-hostedThe repository is public and specifies an MIT license.
WavtokenizerEncode audio into discrete codes for downstream audio-language-model work.Linux, Windows, macOSThe repository is public and lists an MIT license. The README links to model checkpoints and collections on Hugging Face.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
7 November 2026
Original website title
drethage/speech-denoising-wavenet · Official homepage title · retained in its source language
How we review sourcesSuggest a correction