Together AI TTS runs $4 to $65 per million characters; Sume is $47.50

Together AI's pricing page lists text-to-speech models from $4.00 to $65.00 per million characters, a 16-fold spread. Sume's TTS is $0.0475 per 1,000 characters, which is $47.50 per million, so it sits inside that range, between the open-model rows and the Cartesia rows.

That spread is the real lesson. The price per character depends more on which voice model you pick than on which API sells it. Read the numbers below as a way to size a budget, not as a verdict on voice quality, which this post does not test.

What does Together AI list?

These are the audio text-to-speech rows from Together AI's pricing page, read on 2026-10-03, with Sume's published rate alongside. The cost column is for 100 scripts of 1,200 characters, which is 120,000 characters.

How does Sume count characters?

Sume's rate card says TTS is priced per transcript character after the house margin, that spaces and punctuation count, and that the maximum is 20,000 characters per job. A 1,200-character script is therefore $0.057. The per-length ladder is in Sume text to speech cost by script length.

The margin is stated on the pricing side: provider list times 1.25. The pages I read do not list which voice models Sume carries, so check the live model list in the dashboard before you budget around a specific voice.

Is Sume cheaper than Together for the same voice?

Not necessarily. Together lists the Cartesia models at $65.00 and Sume's rate is $47.50, but the two are not the same product: different margin, character limits, output handling and delivery. I did not confirm that Sume carries Kokoro or Orpheus, so do not read the table as a claim that Sume offers a $4 or $15 route. It does not publish one.

What decides the choice?

Voice quality and language come first, then limits, then price. At 120,000 characters the whole spread is $0.48 to $7.80, so for most video work narration is a rounding error next to the picture. The bill only matters at audiobook or call-centre scale: 20 million characters is $950 on Sume, $1,300 at $65 and $80 at $4.

For video, spend your attention on the per-second rate of the visual model, which dominates. A 30-second Avatar Plus clip is $7.35 before you count the voice.

What would change this comparison?

A different default model on either side, a new margin, or a volume tier. Both pages are dated, so re-read them before you commit a large budget.