Skip to main content
Amazon SageMaker
Voice Agent Unit Economics

Amazon Polly Pricing Calculator

Compare Standard, Neural, and Generative voice tiers — size TTS cost per user session before you ship voice features.

  • No signup
  • Instant results
  • ~2 min

Rates reviewed

Step 1 of 250%

Voice tier & volume

us-east-1list rates, pricing as of 2026-07-06. Speech marks billed separately. SSML and plain text use the same per-character rates. Source: AWS public pricing.

chars

Average voice response ≈ 500–2,000 characters.

sessions

Who This Tool Is For

AI product teams and voice/GenAI builders sizing text-to-speech before shipping conversational features.

Why We Built This Tool

Neural voices cost 4× Standard — a voice agent doing 5M characters monthly jumps from $20 to $80 before speech marks and Bedrock inference.

What Problem It Solves

  • Tier mismatch. Generative or Long-Form on short chat responses wastes 5–25× per character vs Neural.
  • Response length. Verbose LLM outputs inflate TTS character counts — cap responses before synthesis.
  • Speech marks. Viseme and word-timing metadata doubles character-metered charges when enabled.

See Transcribe pricing or GenAI on AWS.

Frequently Asked Questions

Which Polly tier should I use for a voice agent?

Neural is the production default for conversational agents. Standard works for internal tools and alerts. Generative fits expressive, character-driven delivery. Long-Form is for audiobook-length narration only.

Is there a Polly free tier?

Amazon Polly offers 5M characters per month free for the first 12 months on Standard and Neural voices in supported regions. This calculator models post-free-tier production rates.

How do speech marks affect cost?

Speech marks (visemes, word boundaries) bill at $4.00 per 1M characters in addition to synthesis charges when requested.

Polly vs Bedrock for voice output?

Polly is purpose-built TTS with predictable per-character pricing. Bedrock multimodal models can generate audio but bill per token — compare total voice-agent cost with both calculators.

How many characters is a typical voice response?

Short agent replies run 200–800 characters. A 2-minute narration at ~150 words/minute is roughly 1,500 characters.

Where do these prices come from?

AWS published us-east-1 list prices (Amazon Polly pricing), pricing as of 2026-07-06.

100+
AWS engagements
25+
GenAI deliveries
AWS Select
Tier Partner