AI Text-to-Speech (TTS) Cost Calculator

The calculator treats every TTS provider the same way: take the synthesis volume in characters, divide by 1,000,000, and multiply by the vendor's published $/1M-char price. Each price comes from the vendor pricing pages cited at the bottom of this page, hand-verified on 2026-05-12and stored alongside its source URL in the calculator's code.

If you want to hear what these voices actually sound like before committing to a paid API, the free AI Voice Generator renders short text to an MP3 in the browser — useful for sanity- checking pronunciation on Sinhala or Tamil script before you size a monthly bill here. And if AI voice is one line item in a wider model budget, the AI Image Generation Cost Calculator and AI Embedding Cost Calculator use the same USD-and-LKR, sources-cited methodology for the other halves of a typical AI stack.

The core formula is intentionally boring:

chars_per_month  = mode == "characters"
                     ? input_chars
                     : input_minutes × wpm × 5    // 5 chars/word
billable_chars   = apply_free_tier
                     ? max(0, chars_per_month − free_allowance)
                     : chars_per_month
monthly_usd      = (billable_chars / 1,000,000) × price_per_1M
monthly_lkr      = monthly_usd × usd_to_lkr_rate

The chars-per-word constant (5) follows the standard English measurement (Mayzner & Tresselt 1965: ~4.79 chars/word plus one space, rounded up) used by every major TTS vendor's own cost-estimator example. Override the words-per-minute input (default 150) to match your style — slow narration ≈ 130, conversational ≈ 150, fast ad-read ≈ 200.

ElevenLabs prices in credits, not characters. Each model publishes a credits-per-character rule (Multilingual v2 = 1 credit/char; Turbo v2.5 and Flash v2.5 = 0.5). For a given monthly volume, the calculator walks the published subscription plans (Free → Starter → Creator → Pro → Scale → Business) and picks the cheapest one whose monthly credit pool fits. The reported monthly cost is the plan's flat price; the "$/1M chars" column shows the effective rate on that plan:

effective_$/1M = plan_monthly_usd
                 ÷ (plan_credits ÷ credits_per_char)
                 × 1,000,000

Where a provider has no model in a given tier (Deepgram has no HD variant; OpenAI has no $4/1M budget tier; PlayHT bundles premium with voice cloning), the row is greyed out with a tooltip explaining why — never replaced with a fabricated number.

Free monthly allowances are jurisdiction-aware: Google's tier is permanent; AWS Polly's is first-12-months only on a new AWS account; Azure's F0 allowance is permanent; ElevenLabs Free is 10,000 credits/month forever. The toggle in the calculator lets you compare with and without these allowances applied.

Most Sri Lankan builders pay these bills in USD off a rupee income, so the LKR column matters as much as the USD one. If you earn in dollars, the LKR exchange rate page shows today's dollar rate, so you can see what a given TTS bill really costs in rupees, and if your scripts are in Sinhala or Tamil the Sinhala & Tamil AI Cost Calculator accounts for the extra characters those scripts consume compared to English at the same word count.

Minutes ↔ characters round-trip self-test passes — the conversion is reversible at integer minute inputs.

Worked-example self-test (computed live on this page) — each line reconciles the formula above with the hand-derived numbers in the code header:

• ✓A · 800k chars · standard · OpenAI tts-1 → expected $12.00, got $12.00
• ✓A · 800k chars · standard · Google Neural2 (free covers) → expected $0.00, got $0.00
• ✓A · 800k chars · standard · Azure Neural Std (0.5M free) → expected $4.80, got $4.80
• ✓B · 250k chars · premium · OpenAI tts-1-hd → expected $7.50, got $7.50
• ✓B · 250k chars · premium · Polly Generative (100k yr1 free) → expected $4.50, got $4.50
• ✓C · 0 chars · standard · any provider → expected $0.00, got $0.00
• ✓D · 10M chars · budget · Google Standard (4M free) → expected $24.00, got $24.00
• ✓D · 10M chars · budget · Polly Standard (5M yr1 free) → expected $20.00, got $20.00
• ✓E · boundary 1,000,000 chars · Google Neural2 (1M free) → expected $0.00, got $0.00
• ✓F · 1,000,000,000 chars · standard · OpenAI tts-1 → expected $15000.00, got $15000.00
• ✓G · minutes mode · 30min × 150wpm = 22,500 chars · OpenAI tts-1 → expected $0.34, got $0.34
• ✓H · 250k chars · premium · ElevenLabs Multilingual v2 (Pro plan fits) → expected $99.00, got $99.00
• ✓I · 250k chars · standard · ElevenLabs Turbo v2.5 (Pro plan fits) → expected $99.00, got $99.00

Out of scope for v1: live vendor price scraping (brittle and ToS- restricted; quarterly hand-verification is more honest); voice quality benchmarks (this tool prices, it doesn't rank); custom / cloned voice training cost (bespoke per-vendor pricing); real-time streaming surcharges; self-hosted GPU TTS rental (Coqui XTTS, Piper, Kokoro on RunPod — different unit economics); music or sound-effect generation (Suno, Udio, ElevenLabs Sound Effects); and speech-to-text pricing — the sibling AI Audio Transcriber handles that side, in-browser.

• OpenAI — API Pricing (tts-1 $15/1M, tts-1-hd $30/1M chars)
• ElevenLabs — Pricing (Free, Starter, Creator, Pro, Scale, Business plans)
• ElevenLabs — Text-to-Speech API reference (per-model credit rules)
• Google Cloud Text-to-Speech — Pricing (Standard, Neural2/WaveNet, Studio, Chirp 3 HD)
• Azure AI Speech — Pricing (Neural Standard, Neural HD)
• Amazon Polly — Pricing (Standard, Neural, Long-form, Generative + AWS Free Tier)
• Deepgram — Pricing (Aura TTS)
• Cartesia — Pricing (Sonic Turbo, Sonic 2)
• PlayHT — Pricing (Creator $39/mo, Unlimited $99/mo)
• Central Bank of Sri Lanka — Daily indicative USD→LKR rates

Pricing was last cross-checked against each vendor's pricing page on 2026-05-12. The table is reviewed quarterly and whenever a vendor announces a tier change. Spotted a stale row? Email the address below — most fixes ship within 24 hours.