Guide

Turn the podcasts you already listen to into vocabulary practice

Podcasts are some of the best language input there is: hours of natural speech, in your ears, on your commute. They're also the easiest to forget — listening is passive, and the phrase you half-caught at a red light is gone by the time you park.

Here's how to keep what you hear.

Listening needs a recall step

Comprehensible input works — but input alone builds recognition, not recall. To actually own a word you have to be asked to produce it, more than once, spaced out over time. With a podcast that's hard: you can't easily pause-and-mine audio the way you can a subtitled video, so most people just listen and hope.

The fix is to give the episode a second life as review: pull the useful language out of the audio and turn it into short questions you get asked again later.

Doing it with Langcap

Drop an audio file in — an mp3, an m4a, a downloaded episode — and Langcap transcribes it on your machine (Whisper, on your GPU), then mines it into bite-sized questions, each stitched back to the exact second in the audio it came from. A spaced-repetition scheduler brings each word back as it fades, and you can Play the episode back with the words highlighting as they're spoken, or drop into Recall — the question loop scoped to that one episode.

Everything runs locally: the audio never leaves your machine, no account, and it works offline after the one-time model download.

A note on languages

Audio is the on-ramp that works for every language Langcap supports, including Japanese and Korean — so podcasts and audio are a great way to study those two today, while their text read-aloud voice is still coming. The full set of languages with both text and audio on-ramps is English, Spanish, French, German, Italian, Portuguese, and Chinese. It's a young Windows beta and needs a Vulkan-capable GPU.