Gemini 3.8 text-to-speech models now available on AI Gateway
Source
vercel.com
Author
Zachary Chen
Date
Context
Vercel's AI Gateway announcement adds Google's Gemini 3.8 Flash and Flash-Lite speech-generation models. Applications send text and receive audio through the gateway, which also tracks usage and cost.
The choice concerns how much control the application needs over the performance. Vercel positions Flash-Lite for high-volume generation with tone and pacing controls. Flash adds voice and character design through written instructions, including accents and acting cues. Both support long narration and two-speaker dialogue, allowing developers to specify delivery as well as the words to be spoken.
Key takeaways · AI-distilled
Vercel's AI Gateway now serves Google's Gemini 3.8 Flash-Lite TTS and Flash TTS, which both cover 100+ languages with long-form narration, delivery control, and two-speaker dialogue.
Flash-Lite TTS targets high-volume generation with tone, pacing, and line-by-line controls, while Flash TTS adds voice and character design from natural-language prompts, including acting cues and accents.
Both are called through the AI SDK's experimental_generateSpeech with a model id, voice, and output format, and the gateway adds per-request usage and cost tracking, routing rules, and bring-your-own-key.
