Usage calculation specifications for each operation in the Gemini Voices API

Gemini API
1

I would like to clarify a few details regarding the Gemini 3.8 Flash TTS Voices API.

  • In POST /v1beta/voices, how are replicated (created from audio recordings) and prompted (created from text) each billed in terms of SKU and usage metrics?
  • Do consent audio processing, preview audio generation, and voice storage, retrieval, updating, or deletion incur separate usage charges?
  • What are the billing conditions for failures and timeouts, and how can these be reconciled with the billing report?
  • If certain operations do not incur usage, could you point me to official documentation outlining the applicable conditions and effective dates/deadlines?

Billing Support directed me to check here regarding the specifications for how usage is incurred.

Referenced official documentation:

2

Hello @masafumi,

  • Creating voices (POST /v1beta/voices): "prompted" creation is billed at the TTS model’s token rates, and the response’s usage field shows the tokens, including the sample_audio preview (Voices API, Pricing). "replicated" creation reports no token usage. Speech generated with either voice is billed at normal TTS rates.
  • Consent audio, storage, and management: Gemini API charges are token-based (What am I billed for?), and the Pricing page lists no separate charge for consent audio, voice storage, or listing, getting, or deleting voices. There is no update method (Voices API).
  • Failures and billing reports: requests that fail with a 400 or 500 error aren’t charged for tokens but still count against your quota (Am I charged for failed requests?). To reconcile, compare usage values with Cloud Billing Reports grouped by SKU (Evaluate your billing account spend).