Humanness Index™ · TTS model

ElevenLabs Eleven v4

by ElevenLabs

Eleven v4 is the quality focused member of the v4 family, released on September 28, 2026.

Rank: #13
Humanness: 71
Likely rank: Provisional
Blind votes: 0

Standings as of Oct 27, 2026, 19:01 UTC

A real arena clip: a cloned source voice reading a customer support prompt at phone quality.

Eleven v4 key stats

Latency (measured): 671 ms¹
Languages: 90+²
Price / 1M chars: $80³
Streaming: Yes⁴
Voice cloning: Instant voice cloning⁵
Released: September 28, 2026⁶

Vapi streaming benchmark (50 trials per model) (checked 2026-10-27) Median of 50 sequential Text to Dialogue WebSocket trials after two discarded warm ups, October 27, 2026; PCM 24 kHz; Clara v4 Instant Voice Clone; stability 0.35, similarity 0.95. Includes network RTT, excludes WebSocket connection setup.
elevenlabs.io/docs/overview/models (checked 2026-10-27)
elevenlabs.io/pricing/api (checked 2026-10-27) Published ElevenAPI rate of $0.08 per 1K characters, normalized to $80 per 1M. Enterprise contract rates may differ.
elevenlabs.io/docs/api-reference/text-to-dialogue/ttd-websocket (checked 2026-10-27)
elevenlabs.io/v4 (checked 2026-10-27) The benchmark uses fresh Instant Voice Clones from 30 second licensed reference samples, rather than the earlier models' professional clones.
elevenlabs.io/blog/eleven-v4 (checked 2026-10-27)

Background

Eleven v4 is the quality focused member of the v4 family, released on September 28, 2026. ElevenLabs describes a new architecture with greater control over emotional delivery, more reliable speaker identity, and support for more than 90 languages. The Index evaluates its plain English scripted output using the same licensed source speakers as every other model.

Sources: elevenlabs.io, elevenlabs.io

On the Index

The arena clips use four fresh Instant Voice Clones made from 30 second excerpts of the licensed Clara, Emma, Godfrey, and Nelliot recordings. Each clone reads all 20 frozen prompts through the Text to Dialogue WebSocket with stability 0.35 and similarity 0.95. Our October 27, 2026 benchmark measured a median of 671 ms to first audio over 50 sequential PCM streaming trials after two discarded warm ups, using the same Clara clone and including network RTT.

Sources: elevenlabs.io, elevenlabs.io

Position in the rankings

Standings as of Oct 27, 2026, 19:01 UTC

Rank	Provider	Model	Humanness	Latency
#11	ElevenLabs	Flash v2	76	226 ms
#12	Cartesia	Sonic 3.6	71	101 ms
#13	ElevenLabs	Eleven v4	71	671 ms
#14	ElevenLabs	Eleven v4 Turbo	71	158 ms
#15	Murf	Falcon 2.1	71	117 ms

See the full Humanness Index™ rankings

Frequently asked questions

How is Eleven v4 tested on the Humanness Index™?: Listeners hear Eleven v4 against another model in a blind head to head round, both voices reading the same customer support prompt from the same cloned source voice, and they pick whichever sounds more human. Its Humanness score derives purely from those votes.
How does Eleven v4 differ from Eleven v4 Turbo?: Eleven v4 is tuned for produced speech where quality is the priority. Eleven v4 Turbo targets realtime speech generation. Both use the same four cloned benchmark speakers and the same 20 English prompts here; synthesis latency is measured separately from listener votes.

Keep exploring

ElevenLabsAll ElevenLabs models on the Index Turbo v2Rank #10 · Humanness 77 Turbo v2.5Latency 265 ms Flash v2Rank #11 · Humanness 76 Flash v2.5Rank #19 · Humanness 69 Eleven v3Rank #1 · Humanness 98 Multilingual v2Latency 1006 ms Eleven v4 TurboRank #14 · Humanness 71

Back to the Humanness Index™

Find the most human-sounding voice for your agent.

Compare the models in blind tests, read the methodology, or get in touch.

Read the methodology Star on GitHub

Build a TTS model? Add yours to the Index.

Eleven v4 key stats

Latency (measured)

671 ms¹

Languages

90+²

Price / 1M chars

$80³

Streaming

Yes⁴

Voice cloning

Instant voice cloning⁵

Released

September 28, 2026⁶

Vapi streaming benchmark (50 trials per model) (checked 2026-10-27) Median of 50 sequential Text to Dialogue WebSocket trials after two discarded warm ups, October 27, 2026; PCM 24 kHz; Clara v4 Instant Voice Clone; stability 0.35, similarity 0.95. Includes network RTT, excludes WebSocket connection setup.

elevenlabs.io/docs/overview/models (checked 2026-10-27)

elevenlabs.io/pricing/api (checked 2026-10-27) Published ElevenAPI rate of $0.08 per 1K characters, normalized to $80 per 1M. Enterprise contract rates may differ.

elevenlabs.io/docs/api-reference/text-to-dialogue/ttd-websocket (checked 2026-10-27)

elevenlabs.io/v4 (checked 2026-10-27) The benchmark uses fresh Instant Voice Clones from 30 second licensed reference samples, rather than the earlier models' professional clones.

elevenlabs.io/blog/eleven-v4 (checked 2026-10-27)

Background

On the Index

Rank

Provider

Model

Humanness

Latency

#11

ElevenLabs

Flash v2

226 ms

#12

Cartesia

Sonic 3.6

101 ms

#13

ElevenLabs

Eleven v4

671 ms

#14

ElevenLabs

Eleven v4 Turbo

158 ms

#15

Murf

Falcon 2.1

117 ms

Frequently asked questions

How is Eleven v4 tested on the Humanness Index™?

Listeners hear Eleven v4 against another model in a blind head to head round, both voices reading the same customer support prompt from the same cloned source voice, and they pick whichever sounds more human. Its Humanness score derives purely from those votes.

How does Eleven v4 differ from Eleven v4 Turbo?

Eleven v4 is tuned for produced speech where quality is the priority. Eleven v4 Turbo targets realtime speech generation. Both use the same four cloned benchmark speakers and the same 20 English prompts here; synthesis latency is measured separately from listener votes.