Introducing Eleven v4, our most emotive model
Short summary
ElevenLabs released Eleven v4, a text-to-speech model that generates emotionally expressive speech by interpreting tone, pacing, and context. It ranks #1 in independent benchmarks and was preferred by ~75% of listeners in blind tests against competitors.
- Eleven v4 can generate speech with distinct emotions (dramatic, tender, urgent, comedic) while maintaining speaker identity, and users can control delivery through natural language prompts and inline tags like [laughs] or [said angrily in French accent]
- Eleven v4 Turbo achieves ~150ms latency for conversational agents, solving the traditional speed-vs-quality tradeoff that forced developers to choose between fast but robotic or slow but expressive voices
- The model supports 90+ languages and can clone voices that speak fluently in any language while adopting native accents and retaining the original voice's identity, improving dubbing and localization work
Full summary
ElevenLabs released Eleven v4, a text-to-speech model that handles emotional nuance and context in ways previous systems couldn't. The key difference: it interprets tone, pacing, and character from text rather than just reading words mechanically. You can now use inline tags like [laughs] or [said angrily in French accent] and the model actually follows them, plus it handles International Phonetic Alphabet phonemes more reliably. In blind tests, listeners preferred it about 75% of the time over competing models. They also released Eleven v4 Turbo, a faster variant with roughly 100-150ms latency—fast enough to feel responsive in live conversations. The model preserves individual voice identity better across long-form content, so characters and narrators stay consistent rather than drifting.
The multilingual performance is where things get interesting. A single voice can now speak 90+ languages while adopting native-speaker accents without losing its original identity. Previously voices would drift back toward their source accent partway through. This matters for dubbing and localization—you keep your brand voice or actor across markets without sounding like someone doing an accent. Voice cloning improved too: you need just 10 seconds of audio for an Instant Voice Clone, and the system now supports Professional Voice Clones for higher-fidelity work. Request stitching (chaining generations together for longer content) is significantly more reliable, which improves the practical experience of actually using their tools.
Eleven v4 Turbo specifically targets conversational agents and low-latency use cases. The problem it solves is real: companies had to choose between fast agents that sounded robotic or expressive ones that felt sluggish. Turbo combines speed with emotional range, which opens applications in customer service, gaming, and healthcare where response time and tone both matter. The company built Turbo and their ElevenAgents platform as a unified system rather than stitching together separate models, which they claim produces more reliable outputs than competitors piecing together different vendors' tech. Both models are available now through their API, creative tools, and agent platform.