Skip to content

Benchmark

EmergentTTS-Eval Win-Rate

Scores are percentages. Higher is better.

Models measured
9
Scores recorded
13
Sources
1
IQ weight
0.39

Score distribution

53.8%
Mean
60.2%
Median
19.3%
Std dev
25.9%
10th percentile
75.4%
90th percentile
17.1%
Min
17.1%75.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.39 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 Flash TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data75.6%
2Gemini 2.5 Flash TTSvoice: Zephyr· EmergentTTS-Eval data75.4%
3GPT-4o Audio Previewvoice: Ballad, strong-prompt· EmergentTTS-Eval data72.7%
4Gemini 2.5 Pro TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data72.2%
5GPT-4o Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data63%
6GPT-4o mini Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data61%
7GPT-4o mini TTSvoice: Alloy, strong-prompt· EmergentTTS-Eval data60.2%
8GPT-4o Audio Previewvoice: Alloy· EmergentTTS-Eval data58.1%
9GPT-4o mini Audio Previewvoice: Alloy· EmergentTTS-Eval data52.2%
10Speech 02 HDvoice: English_expressive_narrator· EmergentTTS-Eval data36.8%
11Orpheus TTSvoice: Tara· EmergentTTS-Eval data29.4%
12Kokoro 82Mvoice: af_heart· EmergentTTS-Eval data25.9%
13F5 TTSvoice: basic_ref_en· EmergentTTS-Eval data17.1%

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