pronunciation-drilllisted

# Pronunciation Drill ## When to Use Load this skill when: - A video script is locked and the narrator wants to practice difficult words before recording - The narrator has a non-native accent and specific words trip them up - The script contains 5-10 high-frequency or high-stakes words that get mispronounced Do NOT load for: - General English pronunciation tutoring (this is script-specific, not curriculum) - Words that don't appear in the actual script (no value) - More than ~12 words at once (drill becomes too long; split into two skills) ## Composes With - `script-voice` — run this after the script is finalized - `descript-export-flow` — drill audio is separate from the final video pipeline ## Constraints ### 1. Word list comes from the actual script Every word in the drill must appear in the locked script. No "general useful words." Practice has to map directly to what they're about to record. ### 2. Each entry has both a slow word and a script sentence The drill loop per word is **always** `[slow word] → silence → [slow word] → silence → [normal sentence] → long silence`. Don't deviate. ### 3. Use the same voice as the rest of the workflow Default: `af_heart` (Kokoro 82M). This keeps the narrator's reference voice consistent across drill, captions check, and any TTS preview. ### 4. Output is loudnorm-ready The final M4A should be at -14 LUFS so the narrator can drop it into AirPods at the same volume as anything else they're listening to. No raw Kokoro ou