ElevenLabs dialogue continuity vs Sume TTS segments: what differs

Sume TTS has no previous_text, future_text or request-id continuity fields. ElevenLabs added continuity controls to Text to Dialogue, per its changelog (Sep 28) and API reference, read 2026-10-01. Sume instead lets you ask for sentence segments with timing and keep a whole script in one call.

What did ElevenLabs add?

Its reference says previous_text and future_text are each up to 100 characters, and previous_request_ids and next_request_ids take up to 3 ids. These let separately generated takes sound continuous.

What does Sume offer?

POST /v1/tts-1.0/generate takes a transcript up to 20,000 characters, so most scripts fit one call and one take. With timestamps.segmentation set to sentence you get gapless sentence segments, and word timestamps are available too. Voice comes from avatar_id, avatar_handle or voice.id.

How do I keep tone across calls?

Keep the same voice and generation_config (speed 0.6 to 1.5, volume 0.5 to 2, an emotion string up to 64 characters) on every call. Sume's docs do not promise that two calls will join without a seam, so listen at the joins.

Which should I pick?

If you need multi-speaker dialogue stitched from many calls, ElevenLabs documents the tools. For one narrator and a timed script, one Sume call is simpler.