Cheapest TTS API per 1,000 characters, October 2026

On list price alone, a hosted open model such as Kokoro is cheapest at about $0.00065 per 1,000 characters, and Sume's TTS at $0.0475 per 1,000 characters is the most expensive row below. That gap is real, and it is the price of a managed, jobs-based API where the same key also does speech-to-text, music and mixing. Decide by what the gap buys you.

The table

Prices below are converted to one unit. The arithmetic: a per-million price divided by 1,000 gives the per-1,000-character price.

What the number does not tell you

• Quality. Artificial Analysis ranks models by listener preference, and price rank is not quality rank.
• Languages. Speechify's Simba 3.2 is English only on its page; Voxtral lists nine languages. Sume requires you to set language for non-English scripts.
• Plan fees. Speechify's tiers run Free, Starter $10, Pro $99 and Scale $499, so the low per-character price sits behind a plan.
• Licence. Open weights can carry non-commercial terms; see Voxtral weights versus the API.

What Sume's price covers

Sume's rate is list times 1.25 as documented in its catalog. For a 5,000-character script, that is $0.2375. The same Sume key can then transcribe, generate music and mix the audio, with jobs you can fetch by id. Sume's pricing page lists every rate.

Limits

If you synthesize millions of characters of English and only need audio files, the cheaper rows will win on cost. If you need one vendor relationship, job records and a mixing step, compare total workflow cost, not the per-character rate. Vendors change prices, so recheck before you commit.