Spanish audio examples

Why I made this

I wanted to practice Spanish flashcards out loud and get a quick right-or-wrong answer. Clef-omni could judge audio without a full conversation, but it sometimes accepted mañana (“tomorrow”) when I expected manzana (“apple”).

I made these examples to see which differences it misses. The clips are synthetic speech from OpenRouter; “Spoken” is the requested script. Each result is a saved Cloudflare Clef-omni judgment of the audio and expected answer, not the spoken script. P(correct) ≥ 0.5 means correct. Stress-only and double-r differences are excluded. This is not an accuracy estimate.