Text to speech add pause: Flux, HeyGen and Gemini markers vs Sume

To add a pause in text to speech you normally put a marker in the text, and each vendor's marker is different. Sume TTS documents no pause marker in its request, so on Sume you shape pauses with punctuation, the speed control, or by generating separate lines and joining them.

Vendor syntax is from the Deepgram, HeyGen and Gemini pages, read 2026-09-30; Sume facts are from the OpenAPI document and Timeline docs.

What pause markers do other vendors use?

What controls does Sume TTS have?

The request has an optional generation_config with volume, speed and emotion controls, and an optional pronunciation_dict_id. Speed is a multiplier in [0.6, 1.5]. None of these inserts a measured silence, so they change delivery, not gap length. Emotion and speed are covered in TTS with emotion, speed and volume.

How do I get a longer gap on Sume?

Generate each sentence or paragraph as its own TTS job and join the files; Timeline audio joins Sume-hosted audio into one gapless file, which means the join itself adds no gap. For a pause of a known length you can place clips on a Timeline spine with later start times. Timeline's audio.mode value silence declares an output length with no audio file, which is for silent videos rather than a silence clip to splice. See join voiceover parts in one render.

Will punctuation alone work?

Commas, periods and paragraph breaks affect phrasing in most voices, but the Sume docs do not promise a duration for any of them. Listen to the result and adjust, and do not paste another vendor's marker into a Sume request: the text would be read as words.