Text-to-Speech · Multilingual
Lithuanian
Language
Benchmarks
XYNaturalness
higher is better1500 is this language's own field — not comparable across languages.
field averagetied with bestOpenAIgpt-4o-mini-tts1206xAI Groktts1235AzureAI Speech (native)1288Google Chirp 3 HD1418Sonioxtts-rt-v21623Gemini3.8-flash-tts1727ElevenLabseleven_v4_turbo1735Gemini3.8-flash-lite-tts1767
10001200140016001800
Time to first audio
RegionThe pause the caller hears before the agent starts speaking. Measured from US East straight to each vendor, on the study's Lithuanian voices.
AzureAI Speech (native)22ms · 59ms
ElevenLabseleven_v4_turbo284ms · 414ms
Sonioxtts-rt-v2290ms · 322ms
xAI Groktts292ms · 351ms
OpenAIgpt-4o-mini-tts461ms · 626ms
Gemini3.8-flash-lite-tts464ms · 552ms
Gemini3.8-flash-tts1.01s · 1.29s
0ms300ms600ms900ms
p50 p90
Robustness
higher is betterWhether the voice says the right words for a number, date, currency amount or operator.
0.000.250.500.751.00
Cost
lower is better$ per 1M characters · vendor list price, the same in every language