Modality

In AI, a kind of data a model can take in or produce: text, code, images, audio, video, documents. A model that handles more than one is multimodal

Input and output are separate questions. Many models read images and audio but write only text; speech synthesis (TTS) writes audio from text, and transcription reads audio into text. The tech registry records both lists per model (modalitiesIn, modalitiesOut).

In interaction design, a channel through which a person and a system communicate: sight, sound, touch, voice, gesture

Designing for more than one modality, such as captions for audio or voice for a screen, is much of accessibility. Dictation apps such as Wispr Flow add voice as an input modality to every app.