Systems and techniques for acoustic-based linguistically-driven automated text formatting of acoustic input are described herein. An audio sample may be obtained that includes multiple words from human speech. The audio sample may be processed to determine a linguistic relationship of the words from the acoustic properties of the words. Based on the determined linguistic relationships, data such as an acoustical language model is generated to enable an arrangement of the words into a cascade format (e.g., a cascade format that establishes horizontal displacement and vertical displacement among the multiple words). A cascaded text output may be provided from the acoustical language model or other types of related data.