Other languages
Does the same model stay accurate outside English?
Ranking
Score out of 100, higher is better. 64 items per model; the thin line on each bar is the 95% range.
When two models' ranges overlap, this test cannot tell them apart.
By language
Accuracy out of 100 for each language; darker is better. Striped cells are languages the tested version does not support.
| Model | Score | Spanish | French | German | Italian | Portuguese | Hindi | Japanese | Chinese | |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | AnyToSpeech Standard voices8 of 8 languages | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 |
| 2 | Grok TTS8 of 8 languages | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 |
| 3 | AnyToSpeech Premium and Characters8 of 8 languages | 99.9 | 100 | 100 | 100 | 100 | 100 | 98.9 | 100 | 100 |
| 4 | Gemini 2.5 Pro TTS8 of 8 languages | 99.9 | 100 | 100 | 100 | 100 | 100 | 100 | 99.5 | 100 |
| 5 | Qwen3 TTS7 of 8 languages | 99.9 | 100 | 100 | 100 | 100 | 100 | n/a | 99.5 | 100 |
| 6 | Gemini 3.8 Flash-Lite TTS8 of 8 languages | 99.7 | 100 | 100 | 100 | 100 | 98.9 | 98.9 | 100 | 100 |
| 7 | Gemini 3.8 Flash TTS8 of 8 languages | 99.7 | 100 | 100 | 100 | 100 | 100 | 97.8 | 100 | 100 |
| 8 | MiniMax Speech 2.8 HD8 of 8 languages | 99.7 | 100 | 100 | 100 | 100 | 100 | 97.8 | 100 | 100 |
| 9 | Eleven v38 of 8 languages | 99.7 | 100 | 100 | 100 | 100 | 100 | 97.8 | 100 | 100 |
| 10 | MiniMax Speech 2.8 Turbo8 of 8 languages | 99.6 | 100 | 100 | 100 | 100 | 97.8 | 98.9 | 100 | 100 |
| 11 | Kokoro 82M3 of 8 languages | 99.6 | n/a | 100 | n/a | 100 | n/a | 98.9 | n/a | n/a |
| 12 | GPT-4o mini TTS8 of 8 languages | 99.5 | 100 | 100 | 100 | 100 | 98.9 | 96.7 | 100 | 100 |
| 13 | Inworld TTS-28 of 8 languages | 98.9 | 100 | 100 | 100 | 100 | 95.7 | 97.8 | 97.8 | 100 |
| 14 | Eleven Flash v2.58 of 8 languages | 98.8 | 100 | 100 | 100 | 100 | 100 | 98.9 | 92.4 | 99.2 |
| 15 | Eleven Multilingual v28 of 8 languages | 98.7 | 100 | 100 | 100 | 100 | 100 | 100 | 89.2 | 100 |
| 16 | TTS-18 of 8 languages | 97.5 | 94.3 | 100 | 100 | 94.8 | 98.9 | 97.8 | 99.5 | 94.6 |
| 17 | Inworld TTS-1.5 Max8 of 8 languages | 97.2 | 100 | 100 | 100 | 99 | 98.9 | 84.8 | 97.3 | 97.7 |
| 18 | Chatterbox Multilingual8 of 8 languages | 68.6 | 100 | 100 | 100 | 100 | 97.8 | 4.3 | 45.9 | 0.8 |
| 19 | TTS-1 HD8 of 8 languages | 63.5 | 51.4 | 80.6 | 100 | 60.4 | 62 | 60.9 | 0 | 93 |
Recognisers are less exact outside English, so 95 and above means "no real errors". Japanese and Mandarin are scored per character.
What we test and how it is scored
Test. 8 sentences each in Spanish, French, German, Italian, Portuguese, Hindi, Japanese and Mandarin, read with the model's default voice (or its voice for that language when the model requires one).
Scoring. Word error rate per language (character error rate for Japanese and Mandarin). Score = average accuracy across the languages the model supports; the number of supported languages is shown next to it.
Full item list and rules: methodology.
Listen for yourself
Unedited clips from the test, in ranking order, including each model's worst miss where it had one.
Think a score is wrong? Check it.
Every test item, scoring rule and known limit is published, and the audio behind each number is on the page.