OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Kaldi Onnx

Kaldi Onnx is a Python developer tool that converts Kaldi speech-recognition neural-network models into ONNX files for inference. It supports Nnet2 and Nnet3 models, including conversion of RNN and TDNN networks into DAG-like networks. MACE is a separate inference framework identified by the project for running the converted models.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Kaldi Onnx — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a Kaldi speech-recognition neural network needs conversion to ONNX and the intended inference framework supports the model’s Kaldi-specific operators.

Who it is for

  • Developers working with Kaldi speech-recognition neural networks who need ONNX models.
  • Teams preparing Kaldi-derived models for inference through MACE on mobile or Linux devices.

Documented uses

  • Convert a Kaldi Nnet2 or Nnet3 text model into ONNX.
  • Prepare RNN or TDNN network structures as DAG-like networks during conversion.
  • Review an ONNX graph and follow the README’s described Kaldi-versus-MACE output comparison workflow.
Before you choose

Requirements & limits

The converter is described in a public GitHub repository under the Apache-2.0 license.

  • Converter input must be a Kaldi text model; binary models require a preceding Kaldi conversion step.
  • The conversion command needs model-specific context, modulus, chunk-size, subsample-factor and network-type values.
  • The project identifies MACE as supporting Kaldi-ONNX operators; deployment requires an inference framework with Kaldi-specific operator support.
  • The README describes broad, but not universal, Nnet2 and Nnet3 component support and directs users to converter/common.py for the available list.
  • The README describes Android, iOS, Linux and Windows inference through MACE.

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

Convert a Kaldi text model to ONNX

Create an ONNX model from a supported Kaldi Nnet2 or Nnet3 text model.

Before you start

  • Python dependencies installed with pip install -r requirements.txt.
  • A Kaldi text model; binary models must first be converted to text with the relevant Kaldi copy tool.
  • Left context, right context and modulus obtained with nnet-am-info or nnet3-am-info.
  • Conversion parameters including chunk size, subsample factor and Nnet type.

Steps

  1. Install the converter dependencies from requirements.txt.

  2. Prepare a Kaldi text model. For binary Nnet2 or Nnet3 input, use the corresponding Kaldi copy utility; the README specifies --raw=true and --prepare-for-test=true for the Nnet3 preparation example.

  3. Obtain the required context and modulus values using Kaldi’s information utility.

  4. Run the command below with the documented input, output, transition-model, configuration, chunk-size, context, modulus, subsample-factor and network-type arguments.

    Documented command
    python converter/convert.py

  5. Optionally review the resulting ONNX graph with ONNX Model Viewer.

Expected outcome

An ONNX model file and the transition-model and configuration outputs specified in the conversion command.

Guide limitations

  • Only Kaldi text models are accepted as converter input.
  • The README says almost all Nnet2 and Nnet3 components are supported; check converter/common.py for the available components.
  • The README does not detail supported model architectures beyond its stated Nnet2 and Nnet3 component coverage.
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
Kaldi OnnxConvert a Kaldi Nnet2 or Nnet3 text model into ONNX.Android, iOS, LinuxThe converter is described in a public GitHub repository under the Apache-2.0 license.
HyperposeCustomize pose-estimation architectures and datasets.Docker, Linux, macOS, Windows
Multi Task NlpDefine multiple tasks and their data and model settings in a YAML file.Linux, macOS, WindowsThe project is an open-source library under the Apache-2.0 license, distributed through its public GitHub repository.
Evidence and transparency

Sources & review

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