Google Gemini 3.8 Flash Text to Speech is Google's new-generation speech model released in September 2026, built for studio-quality creative content. It reads your text aloud word for word as natural, expressive speech and detects the input language automatically across more than 130 languages. Pick one of 30 prebuilt voices from the dropdown, each with its own character such as bright, firm, warm or informative, and choose the spoken language from the second dropdown (auto-detect by default). Unlike the previous generation, this model treats your text as a verbatim transcript, so do not write stage directions in the body: put them on the first line starting with "Style:", for example "Style: cheerful and friendly" or "Style: whispered urgently", and the model adjusts emotion, pace and tone accordingly. You can also insert angle-bracket tags such as <laugh>, <sigh> or <short pause> inside the text to add a laugh, a sigh or a pause at that exact point. Great for audiobooks, podcasts, e-learning narration and product explainers. For high-volume, fast generation switch to Gemini 3.8 Flash-Lite Text to Speech. Text input is limited to 8,192 tokens and the audio is delivered as MP3. Maximum context window (input and output combined): about 24,576 tokens. Maximum total input per request (including history): 8,192 tokens. Maximum output per reply: about 16,384 tokens (model reasoning, if any, counts toward this limit). Output content formats: Audio Pricing: Medium-high Audio output format: MP3. This module does not support continuous conversations. AI can make mistakes. Please verify important information.