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

Eesen

Eesen is an open-source toolkit for training recurrent neural networks that map speech audio to text using CTC and end-to-end recognition workflows.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Eesen — 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 Eesen when developing a CTC-based speech recognizer and choosing between WFST decoding with lexicons and language models or the TensorFlow branch's character RNN language-model decoding.

Who it is for

  • Developers and researchers building or training speech-recognition systems.
  • Practitioners exploring CTC-based recognition with phoneme or character labels.

Documented uses

  • Train a bidirectional LSTM acoustic model to map speech to text.
  • Build an end-to-end recognizer using example setups with phoneme or character labels.
  • Decode CTC outputs with lexicons and language models through WFSTs.
  • Explore character RNN language-model decoding without a fixed lexicon in the TensorFlow branch.
Before you choose

Requirements & limits

The repository is public and lists the Apache-2.0 license.

  • The WFST approach can incorporate lexicons and language models into CTC decoding.
  • The TensorFlow branch's character RNN language-model decoding does not require a fixed lexicon.
  • Example setups follow Kaldi recipes and conventions; the RNN language-model decoding option is identified with a separate TensorFlow branch.

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

Choose an Eesen speech-recognition setup

Select a documented example-system direction for building an end-to-end speech recognizer.

Before you start

  • The Eesen repository and its asr_egs/ example setups.

Steps

  1. Review the example setups under asr_egs/ and their accompanying results.

  2. Choose a phoneme- or character-label setup, as the examples cover both.

  3. For decoding, use the WFST approach when incorporating a lexicon and language model, or consider the TensorFlow branch’s RNN language-model approach when a fixed lexicon is not required.

Expected outcome

A selected example-system direction and decoding approach for an end-to-end speech-recognition workflow.

Guide limitations

  • The repository description does not provide the individual setup commands or steps to train or run a recognizer.
  • RNN language-model decoding is described as available in a separate TensorFlow branch.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
srvk/eesen · Official homepage title · retained in its source language
How we review sourcesSuggest a correction