Abstract: Systems, methods, and computer storage media are provided for initiating a system-wide voice-to-text dictation service in response to a preconfigured gesture. Data input fields, independent of the application from which they are presented to a user, are configured to at least detect one or more input events. A gesture listener process, controlled by the system, is configured to detect a preconfigured gesture corresponding to a data input field. Detection of the preconfigured gesture generates an input event configured to invoke a voice-to-text session for the corresponding data input field. The preconfigured gesture can be configured such that any visible on-screen affordances (e.g., microphone button on a virtual keyboard) are omitted to maintain aesthetic purity and further provide system-wide access to the dictation service.

Abstract: Aspects of the subject matter described herein relate to paragraph snapping. In aspects, a computing device receives user input regarding a selection of text. If the user input is expansion input, the computing device determines whether a set of one or more paragraph snapping conditions is satisfied. If the set of one or more paragraph snapping conditions is satisfied, the selection is snapped to the paragraph. If the user input is contraction input, the selection is shrunk and the paragraph snapping behavior is turned off for the paragraph until the selection activity is finished or until user input indicates that paragraph snapping behavior is to be re-enabled.