Gemini 3.8 Flash TTS
High-fidelity expressive podcasts, interactive video game dialogues, scripted multi-speaker conversations, and prompt-designed custom character voices.
SWE-bench Verified
64.2%
MMLU-Pro
88.5%
MATH-500
91%
Context Window
128,000 characters input dialogue script
| Specification | Value |
|---|---|
| Developer / Lab | Google DeepMind |
| Parameters Total | Proprietary Multimodal Acoustic Flow Matching Transformer |
| Parameters Active | Dynamic Speculative Prosodic Synthesis Engine on TPU v5e |
| Architecture Type | End-to-End Multimodal Latent Flow Matching + Prompt-Steered Voice Conditioning + Native SynthID Spectral Watermarking |
| License | Proprietary Commercial API (Google AI Studio, Gemini API) |
| Input Pricing (per 1M) | $0.15 |
| Output Pricing (per 1M) | $15.00 |