Transcriptive listening: Attentional practices in WeChat voice-to-text repair

This article examines how twenty users detect, correct, accept, or playfully repurpose voice-to-text errors, comparing voice with text. Using screen recordings, screenshots, and stimulated recall interviews, the analysis reveals how users hold voice and text together. I call this attentional work transcriptive listening: a situated comparison between acoustic memory and algorithmic written form, made when users decide whether the output stands for the utterance. Voice-to-text repair is not a technical inconvenience but a site where the speech–writing relation is managed under algorithmic conditions. The algorithm’s orientation toward standard written Chinese naturalizes a language ideology, yet users’ remedial work turns the gap into situated meaning-making. Under algorithmic pre-writing, integrating speech and writing becomes belated and remedial rather than spontaneous.