Soniox TTS at $0.70 per hour vs per-character TTS: compare

Soniox lists text to speech at $4.00 per 1M input text tokens and $21.50 per 1M output audio tokens, which it equates to about $0.70 per hour of generated speech (Soniox pricing, read 2026-10-02). Sume bills text to speech per character, so you cannot compare the two without a speech rate; the reliable way is to measure your own script.

What Soniox publishes

The launch post (read 2026-10-02) dates the TTS launch to 23 April 2026 and says it covers 60+ languages with streaming generation. It does not state voice counts, length limits or cloning on that page, so this post does not either.

Measure a per-character price in hours

Per hour pricing hides how fast the voice speaks. A slow narrator uses more audio per character, so a per-hour price looks cheaper per word at a fast pace and dearer at a slow one. Sume sidesteps this by charging for the text you send, and the TTS Router catalog exposes list prices per character.

A fair comparison in four steps

Use the same language and the same speed on both sides, or the numbers will not line up.

• Pick one real script and count its characters.
• On Sume, generate that script once and note the character-based charge shown for the job.
• Play the result and note its length in seconds.
• Multiply the Soniox hourly rate by that audio length, then compare it with the Sume cost for the same script.

Caveats

Soniox's figure is an approximation by the vendor, and token pricing can shift with the text mix. Sume's price varies by the model id you select in the router. Neither vendor's page gives a quality comparison, so listen to both before choosing on price alone. For other rates, see Cheapest TTS API per 1,000 characters.