Three-ASR intelligibility audit
Semantic WER on the same 400 generated clips per system · lower is better
Qwen3-ASR
Nemotron 3.5
Whisper large-v3
0%
2%
4%
6%
8%
10%
QWEN
NEMO
WHISPER
Inflect-Micro-v2
2.52%
5.45%
2.73%
Inflect-Nano-v2
2.78%
5.63%
2.65%
KittenTTS Nano · Bruno
2.15%
3.96%
2.17%
KittenTTS Nano · Hugo
2.39%
3.80%
2.11%
Piper Low · Ryan
2.81%
5.51%
2.87%
Piper Low · Danny
2.62%
5.60%
2.55%
Supertonic 3 · M2 · 3-step
3.03%
6.04%
3.22%
Supertonic 3 · M2 · 8-step
2.05%
3.56%
8.08%
Supertonic 8-step / Whisper is an insertion-heavy
ASR outlier; Qwen, Nemotron, and listening disagree.
95% matched-prompt bootstrap intervals. Formatting-equivalent number forms are normalized before scoring.