Kaldi
kaldi-asr.org
- Free plan
- Windows
- macOS
- Linux
- Web
- Android
- Self-hosted
What Kaldi is
Kaldi is a free, open-source C++ toolkit for building speech recognition systems, aimed at researchers and professionals. Its code and scripts cover methods such as linear transforms, MMI, boosted MMI, MCE discriminative training and feature-space discriminative training. Kaldi integrates finite-state transducers through OpenFst and includes a matrix library that wraps BLAS and LAPACK routines. Example recipes help build recognition systems, and downloadable model archives cover speech recognition, speaker identification, speech activity detection, diarization and language modeling. The models may be used for any purpose. Kaldi runs on Linux and macOS, supports cross-compilation for Android and supports WebAssembly for in-browser execution. Windows compatibility is not actively maintained, and example scripts do not work there. The project recommends cloning its GitHub repository for the current development version and following the repository’s INSTALL instructions. Its documentation can be expert-oriented; users can seek help through the kaldi-help mailing list. Kaldi is released under Apache License 2.0, which permits commercial use and redistribution.
Who it suits
Kaldi is intended for speech recognition researchers, researchers-in-training and professionals building recognition systems. It suits users comfortable with expert-oriented documentation and source-based installation.
Good to know, and what to watch
- Apache License 2.0 permits commercial use and redistribution
- Includes recipes and downloadable models for several speech tasks
- Integrates finite-state transducers through OpenFst
- Supports Linux, macOS, Android cross-compilation and WebAssembly
- Runs offline and supports real-time recognition
- Windows compatibility is not actively maintained
- Example scripts do not work on Windows
- Documentation is often written for experts
- Some example scripts use SRILM, which is not free for commercial use
Plans
Every plan and priceKaldi
Open source
Getting started
- Open the Kaldi website
- Clone the current source from the Kaldi GitHub repository
- Install dependencies used by the scripts, including Git, bash, Perl, awk, grep, and make
- Follow the repository's INSTALL file
- Choose model archives or example recipes for your speech project
Limits to check
Kaldi has no paid tiers, but Windows example scripts do not work and Windows compatibility is not actively maintained. Some example scripts use SRILM, which the documentation says is not free for commercial use.
The verdict
Kaldi provides researchers with a broad toolkit, modeling recipes and downloadable speech models under a license that permits commercial use. Installation and use call for technical experience, and Windows support has clear limits.
Questions people ask
How much does Kaldi cost?
Kaldi is free and released under Apache License 2.0. The license permits use and redistribution, including commercially.
Which platforms does Kaldi support?
It runs on Linux distributions and Mac OS X, supports Android cross-compilation, and supports WebAssembly for in-browser execution. Windows builds exist, but example scripts do not work there and compatibility is not actively maintained.
Can Kaldi work offline?
Yes. Kaldi supports offline and real-time recognition, with on-device deployment.
What models are available?
Downloadable archives cover speech recognition, speaker identification, speech activity detection, diarization, and language modeling. The models page says they may be used for any purpose.
How is Kaldi installed?
Clone the current source from the Kaldi GitHub repository and follow the instructions in its INSTALL file.
What the maker publishes
- Android and browser builds
- The project repository says Kaldi supports cross-compiling for Android and WebAssembly for in-browser execution. source
- Android support
- Kaldi supports cross-compiling for Android using Android NDK, clang++, and OpenBLAS. source
- Audience limit
- The project says much of its documentation is written for experts and identifies speech recognition researchers and researchers-in-training as its intended audience. source
- Compute environment
- Kaldi can run on a single Unix-like machine, while its documentation describes a Linux cluster with shared storage and optional Nvidia GPUs as an ideal environment. source
- Download
- The project directs users to clone its GitHub repository to obtain the current development version, which may include unfinished or experimental features. source
- External dependencies
- Installation and example scripts use tools including Git, bash, Perl, awk, grep, and make; some example scripts use SRILM, which the documentation says is not free for commercial use. source
- Finite-state transducers
- Kaldi compiles against the OpenFst toolkit and integrates finite-state transducers into its code. source
- Founded
- 2009 source
- FST integration
- Kaldi integrates with Finite State Transducers and compiles against the OpenFst toolkit as a library. source
- Installation
- Users download the current source by cloning the Kaldi GitHub repository and install it using the instructions in the repository's INSTALL file. source
- Intended audience
- The project says its documentation can be expert-oriented and describes speech recognition researchers and researchers-in-training as its intended audience. source
- License
- Kaldi is released under Apache License 2.0, which permits free use and redistribution, including for commercial purposes. source
- Linear algebra
- Kaldi includes a matrix library that wraps standard BLAS and LAPACK routines. source
- Linux support
- Kaldi runs on any Linux distribution and is best tested on Debian and Red Hat Linux. source
- macOS support
- Kaldi runs on Mac OS X. source
- Modeling recipes
- The project provides example recipes for building speech recognition systems and includes models for ASR, speaker identification, speech activity detection, and diarization. source
- Models
- The site offers downloadable model archives for tasks including speech recognition, speaker identification, speech activity detection, diarization, and language modeling. source
- Model use
- The models page states that its downloadable models may be used for any purpose. source
- Operating systems
- Kaldi is best tested on Debian and Red Hat Linux and can run on Linux distributions, Cygwin, or Mac OS X; Windows compatibility is not actively maintained and example scripts do not work there. source
- Project governance
- Kaldi describes itself as a community project with no single legal owner and says contributors license their work directly under Apache terms. source
- Purpose
- Kaldi is a C++ toolkit for speech recognition intended for speech recognition researchers and professionals. source
- Security and patents
- The legal page says Kaldi's use does not protect users against third-party patent claims, including claims by contributors in circumstances described there. source
- Speech recognition methods
- The project lists code and scripts for standard techniques including linear transforms, MMI, boosted MMI, MCE discriminative training, and feature-space discriminative training. source
- Support
- Users can seek help through the kaldi-help mailing list, while bug reports and feature requests belong in GitHub issues. source
Checked against the maker's pages on 9 Oct 2026.