Skip to content

Benchmark

EmergentTTS-Eval Syntactic Complexity WER

Scores are percentages. Lower is better.

Models measured
9
Scores recorded
14
Sources
1
IQ weight
1.16

Score distribution

1%
Mean
0.88%
Median
0.418%
Std dev
0.65%
10th percentile
1.6%
90th percentile
0.5%
Min
0.5%2.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.16 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 data0.5%
2Kokoro 82Mvoice: af_heart· EmergentTTS-Eval data0.65%
3GPT-4o mini Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data0.72%
4Gemini 2.5 Flash TTSvoice: Zephyr· EmergentTTS-Eval data0.73%
5GPT-4o mini TTSvoice: Alloy, strong-prompt· EmergentTTS-Eval data0.84%
6Speech 02 HDvoice: English_expressive_narrator· EmergentTTS-Eval data0.84%
7Gemini 2.5 Pro TTSvoice: Zephyr, strong-prompt· EmergentTTS-Eval data0.87%
8GPT-4o mini Audio Previewvoice: Alloy· EmergentTTS-Eval data0.88%
9GPT-4o mini TTSvoice: Alloy, baseline· EmergentTTS-Eval data1%
10GPT-4o Audio Previewvoice: Ballad, strong-prompt· EmergentTTS-Eval data1.2%
11GPT-4o Audio Previewvoice: Alloy· EmergentTTS-Eval data1.3%
12GPT-4o Audio Previewvoice: Alloy, strong-prompt· EmergentTTS-Eval data1.3%
13Orpheus TTSvoice: Tara· EmergentTTS-Eval data1.6%
14F5 TTSvoice: basic_ref_en· EmergentTTS-Eval data2.1%

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