Text-to-Speech · Multilingual

Japanese

Language

Request a new language

Tell us which language you want added to this benchmark.

Benchmarks

X
Y
200ms400ms600ms800ms120014001600Time to first audio lower is betterNaturalness higher is betterGemini 3.1-flash-tts-previewElevenLabs eleven_v3Cartesia sonic-3.5Inworld tts-2Soniox tts-rt-v2OpenAI gpt-4o-mini-ttsHume octave-2speko.ai/benchmarks

Naturalness

higher is better

1500 is this language's own field — not comparable across languages.

field averagetied with bestHumeHumeoctave-21228OpenAIgpt-4o-mini-tts1338Google Chirp 3 HDGoogle Chirp 3 HD1407MiniMaxMiniMaxspeech-2.6-hd1441Sonioxtts-rt-v21447AzureAzureDragonHDLatestNeural1539InworldInworldtts-21597Cartesiasonic-3.51600ElevenLabseleven_v31646GeminiGemini3.1-flash-tts-preview1756
10001200140016001800

Time to first audio

lower is better

The pause the caller hears before the agent starts speaking. Figures are the English measurement — Japanese has no sweep of its own yet.

Cartesiasonic-3.5108ms · 234ms
InworldInworldtts-2112ms · 125ms
Sonioxtts-rt-v2406ms · 462ms
ElevenLabseleven_v3474ms · 622ms
HumeHumeoctave-2549ms · 637ms
OpenAIgpt-4o-mini-tts724ms · 1.03s
GeminiGemini3.1-flash-tts-preview810ms · 987ms
0ms300ms600ms900ms

p50 p90

Cost

lower is better

$ per 1M characters · vendor list price, the same in every language