Skip to content

Benchmark

EmergentTTS-Eval Foreign Words Win-Rate

Scores are percentages. Higher is better.

Models measured
9
Scores recorded
13
Sources
1
IQ weight
0.32

Score distribution

51.3%
Mean
63.4%
Median
24.6%
Std dev
13.2%
10th percentile
71.6%
90th percentile
3.2%
Min
3.2%73.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.32 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
1GPT-4o mini Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data73.4%
2GPT-4o Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data71.6%
3GPT-4o Audio Previewvoice: Ballad, strong-prompt· EmergentTTS-Eval data70%
4GPT-4o Audio Previewvoice: Alloy· EmergentTTS-Eval data68.8%
5Gemini 2.5 Flash TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data65.6%
6GPT-4o mini Audio Previewvoice: Alloy· EmergentTTS-Eval data64.8%
7Gemini 2.5 Pro TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data63.4%
8Gemini 2.5 Flash TTSvoice: Zephyr· EmergentTTS-Eval data63%
9GPT-4o mini TTSvoice: Alloy, strong-prompt· EmergentTTS-Eval data63%
10Speech 02 HDvoice: English_expressive_narrator· EmergentTTS-Eval data32%
11Orpheus TTSvoice: Tara· EmergentTTS-Eval data14.6%
12Kokoro 82Mvoice: af_heart· EmergentTTS-Eval data13.2%
13F5 TTSvoice: basic_ref_en· EmergentTTS-Eval data3.2%

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