ElevenLabs releases Eleven v4 and real-time v4 Turbo,TTS ranks first in blind testing
BIBIBI
AT A GLANCE
ElevenLabs said that Eleven v4 ranked first in the TTS blind test rankings and introduced a Turbo version with approximately 100 milliseconds of latency.
Article
ElevenLabs reaches the top spot in v4 TTS rankings, surpassing Gemini and Qwen
Beating AI News Flash:ElevenLabs has released the next-generation voice models Eleven v4 and the real-time Eleven v4 Turbo. v4 focuses on more natural emotion and voice control, while Turbo targets real-time voice Agents, with a median inference latency of approximately ms.
In the latest Artificial Analysis TTS blind test rankings, Eleven v4 currently ranks first with 1315 Elo. Cartesia Sonic 3.6 ranks 1275,Gemini 3.8 Flash TTS ranks 1267,Qwen-Audio-3.0-TTS-Plus ranks 1258. The previous-generation Eleven v3 currently has only 1169, ranking 17.
v3 already supports audio tags such as laughter and whispers; v4 mainly makes this control more precise. Users can directly specify tone, pacing, emotion, and sound effects, or describe in natural language how a sentence should be spoken. The model has also expanded language coverage from 70 languages to 90 languages, Instant Voice Clone requires only approximately 10 seconds of audio, and consistency of voices in long-form text and multi-speaker conversations has been improved.
Turbo specifically addresses the speed of real-time conversations. Official documentation shows that the median inference latency of v3 Conversational is approximately ms, while v4 Turbo reduces it to approximately ms.
Beating AI reported ElevenLabs released Eleven v4 and the real-time Eleven v4 Turbo.
02
The report cited the latest Artificial Analysis TTS blind test rankings, saying that Eleven v4 had 1315 Elo and ranked first.
03
In the rankings, Cartesia Sonic 3.6 ranked 1275 Elo,Gemini 3.8 Flash TTS ranked 1267 Elo,Qwen-Audio-3.0-TTS-Plus ranked 1258 Elo。
04
The previous-generation Eleven v3 had 1169 Elo, ranking 17。
05
Eleven v4 focuses on more natural emotion and voice control, while Turbo targets real-time voice Agents.
06
The report mentioned that Eleven v4 Turbo has a median inference latency of approximately 100 milliseconds, while v3 Conversational has approximately 280 milliseconds.
07
Eleven v4 expands language coverage from 70 languages to 90 languages, and Instant Voice Clone requires approximately 10 seconds of audio.
08
The report also said that Eleven v4 improves voice consistency in long-form text and multi-speaker conversations.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
TTS is a technology that converts text into speech. The report said that Eleven v4 received the highest score in a TTS blind test; the Turbo version focuses on reducing wait times in real-time conversations.
Why it matters to readers
If the rankings and latency data in the report are accurate, Eleven v4 may improve voice naturalness and controllability, while the Turbo version may be better suited to real-time voice Agents that require rapid responses.
This indicates that voice AI technology is pursuing goals at the same time: sounding more human and responding faster.
Risks and unknowns
The evidence did not include an ElevenLabs official announcement, product documentation, or pricing page.
The evidence did not include the original Artificial Analysis blind test rankings page, and the ranking may change.
The evidence does not specify the actual availability, pricing, or stability of the relevant models.
The product capabilities and performance data in the evidence are based on reports and have not yet been independently verified against primary sources.
Related Developments
Loading event timeline…
Related concepts
TTS
This term is not in the glossary yet. Browse related concepts in the glossary.