Fish Audio Docs
Control instructionsFine-grained emotion

Fish Audio S2

Control emotion and delivery with open-domain inline instructions.

S2 fine-grained emotion

Use natural-language instructions in square brackets immediately before the text they affect. A cue applies until the next cue or the end of the sentence.

[whispering] Don't let them hear you.
I don't want to [whispering] go in.

S2 has no fixed vocabulary. Examples include [happy], [sad], [long pause], [laughter], and free-form descriptions such as [warm, reassuring voice].