ElevenLabs debuts Eleven v3 alpha with expressive TTS and audio tag control

ElevenLabs has released an alpha version of its latest text-to-speech model, Eleven v3. This update introduces multi-speaker dialogues, supports over 70 languages, and enables fine-grained vocal control using inline audio tags. Tags like [excited], [shouting], and [whispers] allow users to direct tone and emotion directly within the text, marking a major step toward more human-like audio generation. The update is now available through the ElevenLabs interface, with an 80% promotional discount valid until the end of June. This release primarily targets creators of audiobooks, narrative content, games, and voice-powered applications who require nuanced, dynamic audio. Compared to previous versions, v3 alpha emphasizes expressive control, making synthetic speech more adaptable and less monotonous. The inline tag system is also poised to streamline voice localization and dubbing workflows, which often rely on manually edited audio tracks. While API access is not yet live, ElevenLabs confirmed it is in development. The model’s expressive flexibility positions it to compete more directly with advanced character voice systems in gaming, entertainment, and e-learning. As synthetic voice use expands across industries, Eleven v3 offers a new benchmark in voice realism without the overhead of manual voice direction.