POST /v1/audio serves both speech and music; the model decides which, and the fields you send
differ.
Speech
Sendtext and a voice_name (or a voice id from your workspace). ElevenLabs Multilingual v2 and
Turbo v2.5 cover most languages; v3 understands expressive cues in the text. Long scripts can be
split into segments with different voices for dialogue.
Music
Sendprompt_tags, an array describing style, mood and instrumentation. For a free-text
description, use one string in that array. Lyria accepts energy and tempo controls;
ElevenLabs Music (music_v1 / music_v2) also accepts an instrumental toggle and lyrics
with [Verse] / [Chorus] markers.
character_count for character-priced speech and
duration_seconds for duration-priced music. Voice-specific tips:
the ElevenLabs prompting guide.