Eleven V3 Audio Tags Express Emotional Context in Speech

Eleven V3 uses bracket-style Audio Tags to add rich emotional nuance—tension, warmth, hesitation, relief, and more—to AI speech. Users can control the voice model's emotional delivery moment by moment with tags like [sigh], [excited], and [tired].

This feature goes beyond simple voice acting to implement situation-appropriate Emotional Context. Emotional states can be changed mid-sentence, allowing natural emotional shifts and tonal transitions even within long sentences.

The main tag types are as follows:

• Emotional states: [excited], [nervous], [frustrated], [calm], etc.
• Reactions: [sigh], [laughs], [gulps], [gasps], etc.
• Cognitive beats: [pauses], [hesitates], [stammers], etc.
• Tone cues: [cheerfully], [flatly], [deadpan], etc.

As this is currently in the Research Preview stage, Professional Voice Clones (PVCs) may not be fully optimized for V3. To fully take advantage of V3's capabilities, it is recommended to use an Instant Voice Clone (IVC) or an already-designed voice.