🥊 ASR Arena

Two tiny ASR models transcribe your mic side-by-side. Same runtime, same audio — you judge speed vs accuracy.

first run downloads both models (cached after)

Moonshine Tiny ~27M params

—

  • latency —
  • words/sec —
  • WER —
VS

Whisper Base.en ~74M params

—

  • latency —
  • words/sec —
  • WER —

🏆 Scoreboard (wins by lower WER; tie → lower latency — saved locally)

Moonshine: 0 Whisper: 0