NVDA is a free, open source screen reader for Microsoft Windows. It is developed by NV Access in collaboration with a global community of contributors. To learn more about NVDA or download a copy, visit the main NV Access website.
- NV Access: The main home of NVDA
- NVDA on GitHub
- NVDA issues on GitHub: Bug reports, feature requests, etc.
- NVDA development snapshots: Automatically generated builds of the project in its current state of development
- NVDA add-ons: Get add-ons to enhance NVDA
- Translating NVDA: Information about how to translate NVDA into another language
- NVDA community wiki: Articles contributed by the community
- NVDA Controller Client (2010-02-19): NVDA API for external applications to directly speak or braille messages, etc.
- NVDA Developer Guide
- Contributing to NVDA: Guidelines for contributing to the NVDA source code
- NVDA development email list (archives): Discussion about NVDA development
- NVDA commits email list: Notifications for all commits to the Git repository
The NVDA project uses the Git version control system for its source code and documentation.
The NVDA Git repository is located at https://github.com/nvaccess/nvda.git. You can clone it with the following command, which will place files in a directory named nvda:
git clone --recursive https://github.com/nvaccess/nvda.git
The --recursive option is needed to retrieve various Git submodules we use.
The NVDA source depends on several other packages to run correctly.
The following dependencies need to be installed on your system:
- Python, version 2.7.13, 32 bit
- Microsoft Visual Studio Community 2015 with Update 3:
- To download, you will need to join the Visual Studio Dev Essentials program. After you have joined, Visual Studio Community 2015 with Update 3 will be available on the Downloads tab.
- When installing Visual Studio, you need to enable the following:
- In Visual C++: Common Tools for Visual C++ 2015, Windows XP Support for C++
- Windows and Web Development -> Universal Windows App Development Tools -> Tools (1.4.1) and Windows 10 SDK (10.0.14393)
Most of the dependencies are contained in Git submodules.
If you didn't pass the --recursive option to git clone, you will need to run git submodule update --init.
Whenever a required submodule commit changes (e.g. after git pull), you will need to run git submodule update.
If you aren't sure, run git submodule update after every git pull, merge or checkout.
For reference, the following dependencies are included in Git submodules:
- comtypes, version 0.6.2
- wxPython, version 3.0.2.0
- Python Windows Extensions, build 218
- eSpeak NG, commit 37121600
- Sonic, commit 4f8c1d11
- IAccessible2, version 1.3
- ConfigObj, version 4.6.0
- liblouis, version 3.0.0
- NVDA images and sounds
- System dlls not present on many systems: mfc90.dll, msvcp90.dll, msvcr90.dll, Microsoft.VC90.CRT.manifest
- Adobe Acrobat accessibility interface, version XI
- Adobe FlashAccessibility interface typelib
- txt2tags, version 2.5
- MinHook, tagged version 1.2.2
- SCons, version 2.4.1
- brlapi Python bindings, version 0.5.7 or later, distributed with BRLTTY for Windows, version 4.2-2
- ALVA BC6 generic dll, version 3.0.4.1
- lilli.dll, version 2.1.0.0
- Handy Tech Braille SDK, version 1.4.2.0
- Updated Handy Tech sbsupport.dll and dealers.dat received on 2014-09-09
- pyserial, version 2.7
- HanSoneConnect.dll, version 2.0.0.1
- SyncBraille.dll, version 1.0.0.1
- Python interface to FTDI driver/chip
- Py2Exe, version 0.6.9
- Nulsoft Install System, version 2.51
- NSIS UAC plug-in, version 0.2.4, ansi
- xgettext and msgfmt from GNU gettext
- epydoc, version 3.0.1 with patch for bug #303
- Boost Optional (stand-alone header), from commit 3922965
These dependencies are not included in Git submodules, but aren't needed by most people.
- If you want to be able to use the Handy Tech braille display driver when running from source code, you will need to install the Handy Tech universal driver
- To generate developer documentation for nvdaHelper: Doxygen Windows installer, version 1.7.3:
Before you can run the NVDA source code, you must prepare the source tree. You do this by opening a command prompt, changing to the root of the NVDA source distribution and typing:
scons source
You should do this again whenever the version of comtypes changes or language files are added or changed. Note that if you want to access user documentation from the help menu while running the source version, you will also need to add user_docs to the commandline like so:
scons source user_docs
While simply testing or committing changes, it may be faster usually just doing scons source as user documentation will change each time the revision number changes.
Among other things, preparing the source tree builds the NVDAHelper libraries.
If trying to debug nvdaHelper, You can control various debugging options with the nvdaHelperDebugFlags command line variable. It takes one or more of the following flags:
- debugCRT: the libraries will be linked against the debug C runtime and assertions will be enabled. (By default, the normal CRT is used and assertions are disabled.)
- RTC: runtime checks (stack corruption, uninitialized variables, etc.) will be enabled. (The default is no runtime checks.)
- analyze: runs MSVC code analysis on all nvdaHelper code, holting on any warning. (default is no analysis).
The special keywords none and all can also be used in place of the individual flags.
An example follows that enables debug CRT and runtype checks
scons source nvdaHelperDebugFlags=debugCRT,RTC
Symbol pdb files are always produced when building, regardless of the debug flags.
However, they are not included in the NVDA distribution.
Instead, scons symbolsArchive will package them as a separate archive.
By default, builds also do not use any compiler optimizations. Please see the release keyword argument for what compiler optimizations it will enable.
To start NVDA from source code, run nvda.pyw located in the source directory.
To view help on the arguments that NVDA will accept, use the -h or --help option.
These arguments are also documented in the user guide.
Since NVDA is a Windows application (rather than command line), it is best to run it with pythonw.exe.
However, if during development you encounter an error early in the startup of NVDA, you can use python.exe which is likely to give more information about the error.
A binary build of NVDA can be run on a system without Python and all of NVDA's other dependencies installed (as we do for snapshots and releases).
Binary archives and bundles can be created using scons from the root of the NVDA source distribution. To build any of the following, open a command prompt and change to this directory.
To make a non-archived binary build (equivalent to an extracted portable archive), type:
scons dist
The build will be created in the dist directory.
To create a launcher archive (one executable allowing for installation or portable dist generation), type:
scons launcher
The archive will be placed in the output directory.
To generate developer documentation, type:
scons devDocs
The developer docs will be placed in the devDocs folder in the output directory.
To generate developer documentation for nvdaHelper (not included in the devDocs target):
scons devDocs_nvdaHelper
The documentation will be placed in the devDocs\nvdaHelper folder in the output directory.
To generate an archive of debug symbols for the various dll/exe binaries, type:
scons symbolsArchive
The archive will be placed in the output directory.
To generate a gettext translation template (for translators), type:
scons pot
Optionally, the build can be customised by providing variables on the command line:
- version: The version of this build.
- release: Whether this is a release version.
- This enables various C++ compiler optimizations such as /O2 and whole-program optimization.
- It also instructs Python to generate optimized byte code.
- publisher: The publisher of this build.
- certFile: The certificate file with which to sign executables. The certificate must be in pfx format and contain the private key.
- certPassword: The password for the private key in the signing certificate. If omitted, no password will be assumed.
- certTimestampServer: The URL of the timestamping server to use to timestamp authenticode signatures. If omitted, signatures will not be timestamped.
- outputDir: The directory where the final built archives and such will be placed.
- targetArchitectures: The target architectures that NVDA should support. Possible values are all, x86 and x86_64. This should generally be left as the default.
For example, to build a launcher with a specific version, you might type:
scons launcher version=test1
If you make a change to the NVDA code, you should run NVDA's automated tests. These tests help to ensure that code changes do not unintentionally break functionality that was previously working. Currently, NVDA has only one kind of automated testing: unit tests.
To run the unit tests, first change directory to the root of the NVDA source distribution as above. Then, run:
scons tests
To run only specific tests, specify them using the unitTests variable on the command line.
The tests should be provided as a comma separated list.
Each test should be specified as a Python module, class or method relative to the tests\unit directory.
For example, to run only methods in the TestMove and TestSelection classes in the file tests\unit\test_cursorManager.py file, run this command:
scons tests unitTests=test_cursorManager.TestMove,test_cursorManager.TestSelection