Research should not be limited by the language you speak or the keyboard you type on.
Research still begins at an interface
A research platform can index excellent material and still be difficult to use if the first step assumes a keyboard, fluent English, or comfort with a text-heavy interface.
That is the access problem behind our work with ScholarXIV. Addis AI speech-to-text, text-to-speech, and realtime models will support voice dictation and listening to research insights inside the platform.
Voice changes both sides of the workflow
Dictation lets a researcher form a question in speech. Text-to-speech lets the answer travel back as audio. Those are different capabilities, and both matter if voice is going to be part of the product rather than an input button added at the end.
The useful version is a loop: speak, search, inspect, and listen. Each step has to preserve the language and the meaning of the research task.
Speak
Dictate a research question or note.
Search
Run the question through the research workflow.
Inspect
Review the sources and written response.
Listen
Hear the research insight in a supported language.
Language access is product work
Expanding access is not a translation pass over a finished interface. Speech recognition, pronunciation, latency, source handling, and the shape of the interaction all affect whether the system is usable.
The partnership starts with voice. Over time, the goal is to extend that access across Ethiopian and African languages as the models and product workflows are ready.
What we are building
ScholarXIV is building the research experience. Addis AI is providing the language and voice infrastructure underneath it. That division keeps the work concrete and gives each team a clear system to improve.