# Text-to-Speech (TTS) Text-to-Speech (TTS) is technology that converts written text into spoken audio using speech synthesis models. Modern TTS systems use deep learning to produce highly natural-sounding speech. ## How it works Modern neural TTS typically follows a pipeline: 1. **Text input**: The system receives written text 2. **Linguistic analysis**: Grammar, sentence structure, and phonetics are analyzed 3. **Prosody generation**: Rhythm, pitch, and intonation are determined for natural delivery 4. **Audio waveform generation**: A neural vocoder converts the intermediate representation into audio ## Approaches - **Concatenative synthesis**: Stitches together pre-recorded speech segments. Sounds natural but inflexible - **Parametric synthesis**: Generates speech from statistical models of acoustic features - **Neural synthesis**: Uses deep learning for end-to-end generation. Current state of the art ## Key neural TTS models (historical) - **WaveNet** (Google DeepMind): Directly generates raw audio waveforms via autoregressive deep generative model - **Tacotron 2** (Google): Converts text to mel-spectrograms using an encoder-decoder architecture with attention - **FastSpeech** (Microsoft Research, 2019): Non-autoregressive approach addressing speed limitations of Tacotron 2 ## Modern open-source TTS models - [[Qwen3-TTS]] (Alibaba) - [[VibeVoice]] (Microsoft) - [[LuxTTS]] - Chatterbox ([[Resemble.AI]], MIT license) - Coqui TTS - Bark (Suno) - Fish Speech ## Notable proprietary TTS models - [[Gemini 3.1 Flash TTS]] (Google) — controllable, expressive, inline audio tags ## Notable proprietary TTS models - [[Gemini 3.1 Flash TTS]] (Google) — controllable, expressive, inline audio tags ## Applications - Voice assistants - Accessibility (screen readers) - E-learning and audiobooks - Podcasts and media production - Multilingual content delivery - [[Voice Cloning]] ## References - Wikipedia: https://en.wikipedia.org/wiki/Speech_synthesis ## Related - [[Voice Cloning]] - [[Gemini 3.1 Flash TTS]] - [[Qwen3-TTS]] - [[VibeVoice]] - [[Voice Clone Studio]] - [[Speech-to-Text (STT)]] - [[Resemble.AI]]