Wispr Flow - Don’t type, just speak
Mini toolsFish Speech
By Fish Audio
Fish-Speech is built around a Dual Autoregressive (Dual-AR) design that separates high-level linguistic structure from fast acoustic detail generation, and it is presented as handling complex linguistic cases like polyphonic text and mixed-language inputs by relying on LLM-based text understanding rather than traditional grapheme-to-phoneme rules.
Audio
Gen 4
Overview
Fish-Speech is an open multilingual text-to-speech system that uses large language models for linguistic understanding to generate natural, expressive speech and support voice cloning.
Last updated: April 6, 2026
