---
title: "Find a voice"
description: "Browse supported text-to-speech models and their naturalness evidence in English and in each measured language."
canonical: "https://speko.ai/voices"
last-updated: "2026-10-15"
---

> ## Speko page index
> The complete index of every page on this site is at: https://speko.ai/llms.txt
> Read it before exploring further. It lists the exact Markdown URL for every canonical HTML page.

# Find a voice

Voices

Measured voice models, with the published naturalness evidence behind them.

## Browse the roster

The catalog endpoint did not answer, so this is the bundled snapshot — which carries the voice models but not their per-voice rosters. The boards below are unaffected.

No per-voice roster in this snapshot. The models are listed below.

## What each one scored

Naturalness is measured per language, not inferred from English — the model that wins one study routinely loses another. Every figure is the board's own.

Model Naturalness Elo Synth english board p50 Cost $ / M chars

eleven_v3_conversational elevenlabs:eleven_v3_conversational

1590 266ms $40.0

eleven_v4 elevenlabs:eleven_v4

~1708 844ms $80.0

eleven_v4_turbo elevenlabs:eleven_v4_turbo

~1670 205ms $40.0

gemini-3.8-flash-tts gemini:gemini-3.8-flash-tts

~1642 1.03s ~$15.0

gemini-3.8-flash-lite-tts gemini:gemini-3.8-flash-lite-tts

~1639 497ms ~$10.0

flux deepgram:flux-haley-en

~1550 106ms $45.0

sonic-3.5 cartesia:sonic-3.5

1574 121ms $50.0

sonic-3.6 cartesia:sonic-3.6

~1578 120ms $50.0

tts-rt-v2 soniox:tts-rt-v2

~1606 381ms ~$13.0

simba-3.2 speechify:simba-3.2

1573 85ms $10.0

simba-3.5 speechify:simba-3.5

~1560 not measured $20.0

inworld-tts-2 inworld:inworld-tts-2

1561 116ms $25.0

s2.1-pro fish:s2.1-pro

~1566 185ms $15.0

lightning_v3.1 smallest:lightning_v3.1

1544 173ms $17.5

grok-tts xai:grok-tts

1495 272ms $15.0

Gradium TTS gradium:default

~1592 105ms $57.8

gpt-4o-mini-tts openai:gpt-4o-mini-tts

1424 691ms ~$20.0

speech-2.8-hd minimax:speech-2.8-hd

1431 294ms $100.0

palabra-tts-v1 palabra:auto

~1520 72ms $12.0

Qwen3-TTS Fast nari:qwen3-tts-fast

~1536 64ms $10.0

Qwen3-TTS nari:qwen3-tts

~1510 75ms $5.0

Qwen3-TTS hopper:qwen3-tts

~1406 not measured ~$5.7

paxa-tts-flash-v1 paxa:paxa-tts-flash-v1

~1325 317ms $10.0

paxa-tts-flash-v2 paxa:paxa-tts-flash-v2

~1490 not measured $10.0

eleven_flash_v2_5 elevenlabs:eleven_flash_v2_5

~1494 not measured $40.0

eleven_multilingual_v2 elevenlabs:eleven_multilingual_v2

~1472 not measured $80.0

eleven_flash_v2 elevenlabs:eleven_flash_v2

~1507 not measured $40.0

sonic-3 cartesia:sonic-3

~1551 not measured $50.0

coda rime:coda

~1535 not measured $50.0

mistv3 rime:mistv3

~1428 not measured $30.0

speech-2.6-turbo minimax:speech-2.6-turbo

~1408 not measured ~$60.0

lightning_v3.1_pro smallest:lightning_v3.1_pro

~1447 not measured $19.5

speech-2.8-turbo minimax:speech-2.8-turbo

~1385 not measured $60.0

assemblyai-tts-1 assemblyai:assemblyai-tts-1

not measured not measured not measured

MAI-Voice-2.1 azure:MAI-Voice-2.1

not measured not measured not measured

MAI-Voice-2.1-Flash azure:MAI-Voice-2.1-Flash

not measured not measured not measured

eleven_v3 elevenlabs:eleven_v3

not measured not measured not measured

drama-3 fish:drama-3

not measured not measured not measured

default hamsa:default

not measured not measured not measured

faseeh-v1-preview munsit:faseeh-v1-preview

not measured not measured not measured

No winner: the measured intervals overlap. Latency and price are the English board’s, by model — a model measured only in this study carries neither.

## Voice models and language coverage

Supported voice models and languages with published measurements. Other language support has not been established by these studies.

Model Provider Voices Measured languages

sonic-3.5 cartesia:sonic-3.5

Cartesia —

ar (Arabic) de (German) en (English) es (Spanish) fil (Filipino) fr (French) hi (Hindi) ja (JA) ko (KO) nb (Norwegian) ta (Tamil) te (Telugu) th (TH) zh (ZH)

eleven_v3_conversational elevenlabs:eleven_v3_conversational

ElevenLabs —

ar (Arabic) de (German) en (English) es (Spanish) fil (Filipino) fr (French) hi (Hindi) ko (KO) nb (Norwegian) ta (Tamil) zh (ZH)

gpt-4o-mini-tts openai:gpt-4o-mini-tts

OpenAI —

de (German) en (English) es (Spanish) fil (Filipino) fr (French) ja (JA) ko (KO) lt (LT) nb (Norwegian) th (TH) zh (ZH)

inworld-tts-2 inworld:inworld-tts-2

Inworld —

ar (Arabic) de (German) en (English) es (Spanish) fil (Filipino) fr (French) ja (JA) ko (KO) th (TH) zh (ZH)

simba-3.5 speechify:simba-3.5

Speechify —

ar (Arabic) de (German) en (English) es (Spanish) fil (Filipino) fr (French) hi (Hindi) nb (Norwegian) ta (Tamil) te (Telugu)

tts-rt-v2 soniox:tts-rt-v2

Soniox —

ar (Arabic) de (German) en (English) es (Spanish) fil (Filipino) fr (French) ja (JA) lt (LT) nb (Norwegian) th (TH)

grok-tts xai:grok-tts

xAI —

ar (Arabic) de (German) en (English) es (Spanish) fil (Filipino) fr (French) lt (LT) th (TH)

speech-2.8-hd minimax:speech-2.8-hd

MiniMax —

de (German) en (English) es (Spanish) fil (Filipino) fr (French) hi (Hindi) nb (Norwegian)

lightning_v3.1 smallest:lightning_v3.1

Smallest AI —

de (German) en (English) es (Spanish) fr (French)

default gradium:default

Gradium —

de (German) en (English) fr (French)

coda rime:coda

Rime —

en (English) ko (KO)

eleven_v3 elevenlabs:eleven_v3

ElevenLabs —

ja (JA) th (TH)

eleven_v4_turbo elevenlabs:eleven_v4_turbo

ElevenLabs —

en (English) lt (LT)

gemini-3.8-flash-lite-tts gemini:gemini-3.8-flash-lite-tts

Google —

en (English) lt (LT)

gemini-3.8-flash-tts gemini:gemini-3.8-flash-tts

Google —

en (English) lt (LT)

paxa-tts-flash-v1 paxa:paxa-tts-flash-v1

Paxa Labs —

en (English) th (TH)

paxa-tts-flash-v2 paxa:paxa-tts-flash-v2

Paxa Labs —

en (English) th (TH)

auto palabra:auto

Palabra —

en (English)

default hamsa:default

Hamsa —

ar (Arabic)

eleven_flash_v2 elevenlabs:eleven_flash_v2

ElevenLabs —

en (English)

eleven_flash_v2_5 elevenlabs:eleven_flash_v2_5

ElevenLabs —

en (English)

eleven_multilingual_v2 elevenlabs:eleven_multilingual_v2

ElevenLabs —

en (English)

eleven_v4 elevenlabs:eleven_v4

ElevenLabs —

en (English)

faseeh-v1-preview munsit:faseeh-v1-preview

Munsit —

ar (Arabic)

flux-haley-en deepgram:flux-haley-en

Deepgram —

en (English)

lightning_v3.1_pro smallest:lightning_v3.1_pro

Smallest AI —

en (English)

mistv3 rime:mistv3

Rime —

en (English)

qwen3-tts hopper:qwen3-tts

Hopper —

en (English)

qwen3-tts nari:qwen3-tts

Nari —

en (English)

qwen3-tts-fast nari:qwen3-tts-fast

Nari —

en (English)

s2.1-pro fish:s2.1-pro

Fish Audio —

en (English)

simba-3.2 speechify:simba-3.2

Speechify —

en (English)

sonic-3 cartesia:sonic-3

Cartesia —

en (English)

sonic-3.6 cartesia:sonic-3.6

Cartesia —

en (English)

speech-2.6-turbo minimax:speech-2.6-turbo

MiniMax —

en (English)

speech-2.8-turbo minimax:speech-2.8-turbo

MiniMax —

en (English)

assemblyai-tts-1 assemblyai:assemblyai-tts-1

AssemblyAI —

not measured

drama-3 fish:drama-3

Fish Audio —

not measured

MAI-Voice-2.1 azure:MAI-Voice-2.1

Azure —

not measured

MAI-Voice-2.1-Flash azure:MAI-Voice-2.1-Flash

Azure —

not measured

## Take one to the API

Pin a voice route with the provider and model in an explicit`routing` object, then send the provider's voice id in`voice`. Router returns raw audio in the encoding declared by the request.

```bash
curl https://router.speko.dev/v1/tts/speech \
  -H "Authorization: Bearer $SPEKO_API_KEY" \
  -H "Idempotency-Key: tts-$(uuidgen)" \
  -H "Content-Type: application/json" \
  -d '{
    "routing": {"mode":"auto","objective":"balanced"},
    "voice": "",
    "input": "Hello from Speko.",
    "audio": {"encoding": "pcm_s16le", "sample_rate_hz": 24000, "channels": 1}
  }'
```

Use `routing.mode: "auto"` to let Router select the voice model by language and objective instead of pinning a route.