Show HN: Deckdrill – Pimsleur-style audio drills from Anki decks, in your voice

Deckdrill converts language-learning Anki cards into audio drills designed for speaking practice. Users can generate cards with an AI provider’s key or import an existing .apkg deck, then choose neural voices from ElevenLabs or Azure, a cloned version of their own voice, or voices built into their device. Drills can be played as long audio tracks and exported as MP3s; the same audio can be added to Anki cards. A Mac and Windows app can save audio using local voices without a speech-service key. API keys stay on the user’s device and are forwarded to the selected provider without being stored; users pay those providers directly.

The Pimsleur-style format gives an English cue, the target sentence, a backward build-up in repeatable chunks, and slow and normal-speed versions. It then prompts recall after intervals - one, four, and ten cards later, plus a final pass. A 30-card drill takes about 16 minutes. The shorter Glossika-style option makes two passes through whole sentences, with pauses and a slow repetition, taking about seven minutes for 30 cards. Deckdrill grounds its approach in research supporting spaced review and active recall, while noting that the precise spacing pattern is less important than an early review and longer delays over time. It favors meaningful sentences over rote sound-pattern practice and supports eight target-language varieties, with English cues.