Skip to content

Benchmark

EmergentTTS-Eval Questions WER

Scores are percentages. Lower is better.

Models measured
9
Scores recorded
14
Sources
1
IQ weight
1.15

Score distribution

0.862%
Mean
0.72%
Median
0.489%
Std dev
0.36%
10th percentile
1.6%
90th percentile
0.27%
Min
0.27%1.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.15 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
1Speech 02 HDvoice: English_expressive_narrator· EmergentTTS-Eval data0.27%
2Gemini 2.5 Flash TTSvoice: Zephyr· EmergentTTS-Eval data0.36%
3Gemini 2.5 Flash TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data0.4%
4GPT-4o mini TTSvoice: Alloy, baseline· EmergentTTS-Eval data0.42%
5GPT-4o mini Audio Previewvoice: Alloy· EmergentTTS-Eval data0.55%
6Kokoro 82Mvoice: af_heart· EmergentTTS-Eval data0.56%
7GPT-4o mini TTSvoice: Alloy, strong-prompt· EmergentTTS-Eval data0.66%
8Gemini 2.5 Pro TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data0.72%
9GPT-4o mini Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data0.74%
10GPT-4o Audio Previewvoice: Alloy· EmergentTTS-Eval data1.2%
11GPT-4o Audio Previewvoice: Ballad, strong-prompt· EmergentTTS-Eval data1.4%
12Orpheus TTSvoice: Tara· EmergentTTS-Eval data1.5%
13F5 TTSvoice: basic_ref_en· EmergentTTS-Eval data1.6%
14GPT-4o Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data1.7%

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