Flux TTS expressivity -2 to 2 vs Sume's emotion guide

Deepgram's Flux TTS has a new expressivity control from -2 to 2, where 0 is nominal. Sume TTS 1.0 has no numeric equivalent: generation_config takes volume, speed and an optional emotion guide that is a short string. To get a flatter or more animated read on Sume, describe it in emotion and compare takes.

Deepgram facts are from its changelog entry for the Aug 26, 2026 self-hosted release; Sume facts from the OpenAPI schema and the API reference, read 2026-10-01.

What does Flux TTS expressivity do?

The entry says the control makes a Flux TTS voice more calm or more animated. Negative values are flatter, positive values more animated, and 0 is the nominal setting. It appears in a self-hosted release note, so check where your deployment runs before relying on it.

What does Sume offer instead?

Three optional controls grouped under generation_config. Only two are numeric, and neither is an expressivity scale.

How do I get a calmer or livelier read on Sume?

Pass a plain description such as calm or excited in emotion; the schema calls it an optional guide, so treat it as a suggestion and listen. Change one thing at a time, and use speed for pacing rather than mood. More on the three fields in text to speech with emotion, speed and volume.

const body = {
  transcript: "The launch moved to Thursday.",
  avatar_handle: "@your_avatar",
  generation_config: { emotion: "calm", speed: 0.9, volume: 1.0 },
  mode: "async",
};
console.log(JSON.stringify(body));

Can I map -2 to 2 onto emotion strings?

Only by your own convention, since Sume documents no scale. A table such as -2 as flat, 0 as no emotion field and 2 as animated keeps your app's dial stable, but the same word may land differently on different voices.