Skip to content

Benchmark

EmergentTTS-Eval Complex Pronunciation Win-Rate

Scores are percentages. Higher is better.

Models measured
9
Scores recorded
13
Sources
1
IQ weight
0.36

Score distribution

38.3%
Mean
44.5%
Median
19%
Std dev
11.6%
10th percentile
60.3%
90th percentile
3.9%
Min
3.9%66.5%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.36 for speech models, measured over 7 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Gemini 2.5 Pro TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data66.5%
2Gemini 2.5 Flash TTSvoice: Zephyr· EmergentTTS-Eval data60.3%
3Gemini 2.5 Flash TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data55.7%
4GPT-4o Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data51.8%
5GPT-4o mini TTSvoice: Alloy, strong-prompt· EmergentTTS-Eval data50.2%
6GPT-4o Audio Previewvoice: Ballad, strong-prompt· EmergentTTS-Eval data45.3%
7GPT-4o Audio Previewvoice: Alloy· EmergentTTS-Eval data44.5%
8GPT-4o mini Audio Previewvoice: Alloy· EmergentTTS-Eval data35.3%
9GPT-4o mini Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data31%
10Kokoro 82Mvoice: af_heart· EmergentTTS-Eval data28.4%
11Speech 02 HDvoice: English_expressive_narrator· EmergentTTS-Eval data12.9%
12Orpheus TTSvoice: Tara· EmergentTTS-Eval data11.6%
13F5 TTSvoice: basic_ref_en· EmergentTTS-Eval data3.9%

Data credits: EmergentTTS-Eval (Boson AI), Apache-2.0 (licence)