LUWRAIN

Intellectual pixel-free technologies

For decades, blind and visually impaired users have interacted with computers primarily through screen readers — software components that translate graphical user interfaces (GUI), originally designed for sighted users, into speech or Braille output. While this approach has made modern operating systems formally "accessible", it has never been truly comfortable. A screen reader is, by its nature, an add-on: it tries to describe in words something that was never designed to be described in words. The result is a partial solution that gives blind users the general ability to do the same things as sighted users, but at the cost of efficiency, predictability, and personal comfort.

LUWRAIN proposes a fundamentally different approach. Instead of adapting a visual interface for non-visual use, LUWRAIN is a Java framework for building applications that are non-visual by design. The screen is still used — it is important for users with low vision and for sighted developers — but it plays a supplementary role. The primary channel is speech, and every object on the screen is constructed from textual data only. There are no icons, no colour-coded states, no widgets whose meaning depends on visual layout. Everything that exists in a LUWRAIN application can be expressed as text, and therefore can be spoken.

The Core Conception

The central idea of LUWRAIN is a text-based environment: the screen is divided into rectangular areas (tiles), and each tile is filled with textual data closely associated with speech output. A blind user navigates by keyboard, hearing the content of the currently focused tile through a speech synthesizer. A sighted user — or a user with partial sight — sees exactly the same content rendered as text on the screen, with adjustable fonts and colours.

This duality is essential. The same data structure drives both the visual rendering and the spoken output. There is no translation layer between "what is on the screen" and "what is announced". The two are aspects of the same underlying object.

The conception is implemented as a Java framework, which means that:

• Application developers do not need prior experience in accessibility engineering. They use ready-made Java classes that already handle speech output, keyboard navigation, and screen rendering correctly.
• Applications are distributed as ordinary .jar files.
• The same LUWRAIN instance can run on any operating system with a Java Virtual Machine, but is most naturally deployed as a complete GNU/Linux distribution, especially for Raspberry Pi or Orange Pi .

Intelligent Assistants and Language Models

A more recent and increasingly central part of the LUWRAIN idea is the integration of large language models (LLMs) and intelligent assistants directly into the interface. This is not an afterthought or an external feature — it changes how blind users interact with information.

For a blind user, certain tasks are disproportionately hard: summarising a long document, extracting the relevant fragment from an unstructured web page, rephrasing a message, translating, or asking a question about the contents of a file currently in focus. These tasks are time-consuming when performed by sequential reading through a speech synthesizer. A language model, integrated into the environment, can perform them in seconds.

LUWRAIN treats the assistant as a first-class participant in the interface:

• Any focused object (a text document, a list item, a fragment of message) can be passed to the assistant as context.
• The user can ask questions about that context in natural language, using a dedicated popup or a system-wide shortcut.
• Assistant replies are themselves text and are therefore announced through the standard speech pipeline, with all the usual navigation features (letter-by-letter exploration, repeating a fragment, copying to the clipboard).
• Different assistants and models can be plugged in through a uniform API, so the user is not locked to a single provider.

The assistant complements the text-based conception rather than competing with it: because every object in LUWRAIN is already text, every object is already in a form that a language model can consume.

What LUWRAIN Is Not

LUWRAIN is not a screen reader. It does not try to describe arbitrary GUI windows. It is not a replacement for Windows, macOS, or GNOME for tasks that genuinely require a graphical interface (photo editing, CAD, etc. – pointless for blind users anyway). It is not isolationist: integrated access to web services, intelligent assistants, a plenty of components from Maven Central and command-line utilities remain with you.

LUWRAIN is best understood as a companion environment for the tasks that blind users perform most often — — reading mail, news, books, documents; writing; programming; managing files; interacting with online services — done in a way that is finally designed for them rather than adapted to them.

Who we are?

The LUWRAIN project was initially started in 2012 by Michael Pozhidaev, an Associate Professor of Theoretical Computer Science at National Research Tomsk State University, as a accessible tool for personal use on Linux.