Drama 3

Drama 3 is Fish Audio's speech synthesis model for directing emotion and delivery with natural-language instructions in a script. Bracketed directions control pacing and intent, target words, sounds and pauses, and assign voices to speakers, combining reading with scene-specific performance. Instead of applying one tone to an entire passage, a script can mark individual emphasis and breaks to shape a character's delivery and the rhythm of a conversation. Phoneme tags specify pronunciation, and in-script language changes are supported, separating voice choice from performance direction. The model detail page's detailed specifications and evaluation analysis cover languages, speaker settings, pronunciation and delivery tags, audio output conditions, naturalness and instruction following.