Humanness Index™ · TTS model

Speechify Simba 3.2

Simba 3.2 is Speechify's streaming native flagship, released to the SpeechifyAI API in July 2026 as the recommended model for new English integrations.

Rank: #3
Humanness: 93
Likely rank: #1–8
Blind votes: 542

Standings as of Aug 18, 2026, 11:54 UTC

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

Simba 3.2 key stats

Latency (measured): 428 ms¹
Languages: English²
Price / 1M chars: $10³
Voice cloning: zero-shot, manual approval⁴
Released: July 8, 2026⁵

Vapi streaming benchmark (50 trials per model) (checked 2026-07-30) Measured on the chunked HTTP /v1/audio/stream endpoint at raw PCM output (16-bit, 24 kHz), using an allow-list voice (the arena clones are not registered on simba-3.2); median of 50 sequential trials, July 2026, including network RTT.
docs.speechify.ai/build/guides/concepts/models (checked 2026-07-23) English only at launch; Speechify says multilingual support will land under the same model id.
speechify.ai/text-to-speech-api (checked 2026-07-23) Flat per-character rate by plan: $10 per 1M characters on Starter, $8 on Pro, $6 on Scale; no credit conversion.
docs.speechify.ai/build/guides/concepts/models (checked 2026-07-23) Cloned voices are supported on simba-3.2, but each voice key currently requires manual Speechify approval.
docs.speechify.ai/build/changelog/2026/7/8 (checked 2026-07-23) API availability of simba-3.2 on the speech and stream endpoints.

Background

Simba 3.2 is Speechify's streaming native flagship, released to the SpeechifyAI API in July 2026 as the recommended model for new English integrations. Speechify positions it on expressivity and time to first byte, and it debuted statistically tied for first place on the Artificial Analysis TTS leaderboard. It serves a curated voice allow list, with cloned voices supported behind a manual approval step.

Sources: docs.speechify.ai, speechify.ai

At a glance

The arena clips for Simba 3.2 were rendered by Speechify with the four licensed source voices cloned on its platform, then verified and hosted by the Index team under the same frozen content hash scheme as every other model, the same vendor supplied path used where the pipeline has no API access. Latency is our own measurement: the 50 trial streaming benchmark ran against a Speechify key in July 2026 and returned a 428 ms median, so the figure here is measured rather than vendor supplied.

Sources: docs.speechify.ai

Position in the rankings

Standings as of Aug 18, 2026, 11:54 UTC

Rank	Provider	Model	Humanness	Latency
#1	ElevenLabs	Eleven v3	96	758 ms
#2	xAI	Grok TTS	93	460 ms
#3	Speechify	Simba 3.2	93	428 ms
#4	MiniMax	Speech 2.8	92	325 ms
#5	Canopy Labs	Orpheus	89	—

See the full Humanness Index™ rankings

Frequently asked questions

How is Simba 3.2 tested on the Humanness Index™?: Listeners hear Simba 3.2 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.
Where did the Simba 3.2 arena clips come from?: Speechify rendered the 80 arena clips (four cloned source voices reading the 20 frozen prompts) with simba-3.2 and supplied them to the Index team, who normalized, verified, and hosted them under the frozen content hash scheme. Blind battles and scoring work exactly as for every other model.

Keep exploring

SpeechifyAll Speechify models on the Index

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.

Simba 3.2 key stats

Latency (measured)

428 ms¹

Languages

English²

Price / 1M chars

$10³

Voice cloning

zero-shot, manual approval⁴

Released

July 8, 2026⁵

Vapi streaming benchmark (50 trials per model) (checked 2026-07-30) Measured on the chunked HTTP /v1/audio/stream endpoint at raw PCM output (16-bit, 24 kHz), using an allow-list voice (the arena clones are not registered on simba-3.2); median of 50 sequential trials, July 2026, including network RTT.

docs.speechify.ai/build/guides/concepts/models (checked 2026-07-23) English only at launch; Speechify says multilingual support will land under the same model id.

speechify.ai/text-to-speech-api (checked 2026-07-23) Flat per-character rate by plan: $10 per 1M characters on Starter, $8 on Pro, $6 on Scale; no credit conversion.

docs.speechify.ai/build/guides/concepts/models (checked 2026-07-23) Cloned voices are supported on simba-3.2, but each voice key currently requires manual Speechify approval.

docs.speechify.ai/build/changelog/2026/7/8 (checked 2026-07-23) API availability of simba-3.2 on the speech and stream endpoints.

Background

At a glance

Rank

Provider

Model

Humanness

Latency

ElevenLabs

Eleven v3

758 ms

xAI

Grok TTS

460 ms

Speechify

Simba 3.2

428 ms

MiniMax

Speech 2.8

325 ms

Canopy Labs

Orpheus

—

Frequently asked questions

How is Simba 3.2 tested on the Humanness Index™?

Listeners hear Simba 3.2 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.

Where did the Simba 3.2 arena clips come from?

Speechify rendered the 80 arena clips (four cloned source voices reading the 20 frozen prompts) with simba-3.2 and supplied them to the Index team, who normalized, verified, and hosted them under the frozen content hash scheme. Blind battles and scoring work exactly as for every other model.