← Back to articles

Speech synthesis cost guide

Text-to-speech API pricing: compare native billing units

Text-to-speech turns text into generated audio. A character rate, an audio-output token rate, and a generated-second rate describe different quantities; their prices must not be ranked as if they were interchangeable.

Updated November 7, 2026 30 speech-synthesis routes in this catalog Catalog generated 2026-11-16

Billing units

Three cost formulas, not one ranking

Per character

Cost = submitted characters × the provider's price per character. Check whether spaces, punctuation, and SSML markup count.

Example with a hypothetical $0.02 per 1,000 characters: 100,000 characters cost $2.00.

Per audio token

Cost = generated audio tokens × output-audio token rate. Text-input tokens may be billed separately.

Do not estimate audio tokens from characters or playback minutes without that route's documented conversion.

Per generated second

Cost = billable generated seconds × the applicable duration rate. A 10-minute clip is 600 seconds only if the provider bills actual output duration.

The presence of an audio-token rate beside a per-second rate does not prove that both charges are added together.

Catalog examples

Open the exact route before estimating

These are examples of currently listed billing dimensions, not a cheapest-model list. Route availability, billing conditions, and other charges need provider confirmation.

Meter to inspectProvider routeWhat to verify
Characters of submitted text OpenAI · tts-1 Current source, scope, tier, and whether other meters apply
Generated audio tokens Google · Gemini 3.8 Flash-Lite TTS Current source, scope, tier, and whether other meters apply
Generated audio seconds OpenAI · gpt-4o-mini-tts Current source, scope, tier, and whether other meters apply

Some listed TTS models expose multiple meters, or no normalized native meter at all. A missing price is unknown, not zero. This catalog does not claim complete provider coverage.

Workflow

Estimate one workload without mixing units

  1. Count monthly text characters, or obtain the provider's audio-token or output-duration estimate.
  2. Choose an exact serving route and confirm the current region, pricing tier, voice options, and any extra text-input or cloning fees.
  3. Use the Voice / Realtime Calculator where its inputs match the route's documented meter. Otherwise calculate from the provider's native unit; do not force a conversion.

Text-to-speech is distinct from speech-to-text transcription and two-way real-time voice agents.

Primary references

Provider billing definitions

Provider documentation and catalog rates can change. The catalog is a decision aid, not an invoice; verify the official rate before purchase.