How it works
Audio output per minute = audio tokens per second × 60 × output price ÷ 1,000,000. Text input per minute = input tokens × input price ÷ 1,000,000. Cost per finished minute = (audio + text) × takes. Example: 25 × 60 × US$9.00 ÷ 1,000,000 = US$0.0135 a minute, which matches Google’s own “US$0.00225 per 10 seconds”. Google has announced higher prices for the Gemini 3.8 TTS models from 1 Jan 2027, so both are listed. OpenAI’s older TTS models are marked deprecated and Anthropic has no TTS API, so only Google’s prices are built in; choose “My own prices” for any other service.
Official TTS prices (US$ per 1M tokens, Standard tier)
| TTS model | Text input | Audio output | Per minute of audio |
|---|---|---|---|
| Gemini 3.8 Flash TTS — until 31 Dec 2026 | 0.5 | 9 | 0.0135 |
| Gemini 3.8 Flash TTS — from 1 Jan 2027 | 1 | 18 | 0.027 |
| Gemini 3.8 Flash-Lite TTS — until 31 Dec 2026 | 0.5 | 6 | 0.009 |
| Gemini 3.8 Flash-Lite TTS — from 1 Jan 2027 | 1 | 12 | 0.018 |
| Gemini 2.5 Flash Preview TTS | 0.5 | 10 | 0.015 |
| Gemini 2.5 Pro Preview TTS | 1 | 20 | 0.03 |
⚠ Estimates for information only — not financial advice.
FAQ
How many input tokens is a minute of speech?
It depends on the language and speed. A minute of calm narration is roughly 130–160 English words; Korean scripts often split into more tokens. The input cost is tiny next to the audio output, so a rough figure is fine.
Can I use the voice commercially?
Check the provider’s terms for generated audio, and never imitate a real person’s voice without their permission.
Checked 14 Nov 2026. Google, Gemini Developer API pricing, ai.google.dev/gemini-api/docs/pricing — TTS models, Standard tier, per 1M tokens; “Audio tokens correspond to 25 tokens per second of audio” (checked 14 Nov 2026). OpenAI model pages for tts-1 and gpt-4o-mini-tts show “Deprecated”; the current OpenAI pricing page lists no TTS model.