Per million characters of text. No seats, no tiers to unlock, no minimums. Early-access list prices. Billing is live: pay only for what you use after your free allowance.
100,000 characters, no card.
Kokoro voices. Fast, clear, the cheapest natural voice.
Record your voice or describe one, and use it everywhere.
Your own voice with a mood. One flat price.
A turn is one question and one spoken answer. It includes the language model, the memory and the speech. Your first 200 turns are free. Speech in by file or live over WebSocket. Phone numbers are not included yet.
File upload or live over WebSocket, with automatic end-of-speech detection. Your first 60 minutes are free. Billed per second of audio. No speaker labels or word timestamps yet.
List price per million characters from each provider's own pricing page, checked 8 October 2026 (sources and caveats on the ). The models differ in voices, languages and features, so this compares cost, not quality.
Method: characters per month × list price per million characters. Every request is assumed to be billed (no caching). Free allowances, introductory credits and promotions are not subtracted. Phrase lengths are measured on the sample lines shown on our use-case pages; rows marked ASSUMPTION are scenario inputs, not measurements. Competitor columns are different models; this compares cost only. At these volumes the absolute dollars are small on every provider.
| Scenario | Inputs | Characters / month | Speakvora catalog ($0.70) | Amazon Polly Neural | Google Neural2 | OpenAI tts-1 | ElevenLabs Flash / Turbo |
|---|---|---|---|---|---|---|---|
| Small app with repeated messages | 30 messages averaging 43 characters (measured on our phone-menu sample lines), played 60,000 times | 2,560,000 | $1.79 | $40.96 | $40.96 | $38.40 | $102 |
| Dispensary operational prompts | ASSUMPTION: 120 prompts averaging 45 characters, played 30,000 times | 1,350,000 | $0.94 | $21.60 | $21.60 | $20.25 | $54.00 |
| Delivery notifications | Notification sample lines average 33 characters (measured); 100,000 sent | 3,300,000 | $2.31 | $52.80 | $52.80 | $49.50 | $132 |
| Clinic reception and navigation | Clinic sample lines average 43 characters (measured); 40,000 plays | 1,709,091 | $1.20 | $27.35 | $27.35 | $25.64 | $68.36 |
| Developer: library plays plus live sentences | 10,000 library plays at 43 characters plus 2,000 live sentences of an ASSUMED 80 characters | 587,273 | $0.41 | $9.40 | $9.40 | $8.81 | $23.49 |
Sources and checked date (2026-10-08): , , , . Polly states that you may cache and replay generated speech at no additional cost; if you synthesise each line once and store it, the gap shrinks for every provider that allows caching. Check each provider's terms first. How billing counts characters: billing explained.
| Item | Speakvora native unit | List price | Notes |
|---|---|---|---|
| Catalog playback via API | Characters | $0.70 per 1M | Library hits are billed by length each time |
| Live generation | Characters | $0.70 per 1M (catalog voices) | New sentences are generated, then reused |
| Your own voice | Characters | $6 per 1M | Private audio kept 75 days after last use; stored on second request |
| Styled voice | Characters | $17 per 1M | Private preview; off in production |
| Transcription | Seconds of audio, listed per minute | $0.0015 per minute | 60 minutes free |
| Agent usage | Turns | $10 per 1,000 turns | Beta; 200 turns free |
| Industry packs | One time per pack | $24 to $269 per industry pack; all-industries bundle $399 | Industry packs (47 live across 23 industries today, more being built for every industry) |
| Storage, delivery | - | No separate charge listed | Audio links expire as described in the docs |
| Unit | Who uses it | Why we do not convert |
|---|---|---|
| Characters | Speakvora, Amazon Polly, Google Cloud, OpenAI tts-1, Deepgram Aura, ElevenLabs | Directly comparable per million characters |
| Minutes / hours | Speech to text and agents at all providers | Compared only within the same product type |
| Tokens | OpenAI gpt-4o-mini-tts and transcribe | Audio output tokens do not map reliably to characters |
| Credits | Cartesia (1 per character for Sonic; 3 per second for STT) | Credit prices depend on the plan |
| Monthly subscriptions | ElevenLabs, Cartesia | Included allowance varies by plan |
| One-time fees | Speakvora packs (when published) | No equivalent found at the voice-API providers reviewed |