D-ID V4 expressives sentiment_id vs Sume's emotion string

In D-ID's V4 expressives, you choose the emotional delivery by sending a sentiment_id with an avatar_id to POST /expressives. Sume has no sentiment preset list: in a TTS request you set generation_config.emotion, a free-text string of up to 64 characters, and generation_config.speed, a multiplier in [0.6, 1.5].

D-ID facts are from its V4 quickstart; Sume facts from the OpenAPI schema and Avatar videos, read 2026-10-01.

What is a D-ID sentiment?

The quickstart describes each sentiment as an emotional state, with examples like Professional, Empathetic, Excited, Friendly and Frustrated. Each one comes with its own voice configuration. You list avatars and their sentiments with GET /expressives/avatars, then the video uses the sentiment's configured voice so tone matches the facial expression.

How does Sume express emotion?

On the TTS request, generation_config has optional volume, speed and emotion. The schema describes emotion only as an optional emotion guide for generation, so treat the wording as a hint and listen to the result. For the talking-video path, each scene in a multi-scene plan has its own voice object, and voice.type: "silence" makes a non-speaking beat that needs a duration.

Preset id or free text: what changes?

Which one should I use?

A fixed list is easier to validate and keep consistent across a team. A free string gives you more room but no guarantee about what a given word does, so settle on a short vocabulary of your own and test each term. The practical steps are in AI avatar emotion and speaking speed.

What do I send on Sume?

Send the script and voice selector as usual, add generation_config with the emotion and speed you want, and use the same job flow as any other Sume request: poll status_url and read result_url when result_ready is true.