Introducing Langcap

Study the shows, podcasts, and books you already love

You watched the whole series in Spanish. You finished the podcast on your commute. You read the book. And a week later, almost none of the language stuck.

That's the gap Langcap is built for. Drop content in. Langcap cuts it into questions — short, answerable ones, each stitched back to the exact clip, line, or page it came from — and then it keeps asking, picking what's useful next. You study the material you already chose to spend time with, instead of a generic word list someone else built.

The problem: native content is the best input, and the hardest to keep

Watching and reading real content in your target language is the method that works — but doing it by hand is a second job. You pause, you look a word up, you copy a sentence into Anki, you find a screenshot, you tag it. Sentence mining is real, effective, and tedious enough that most people quit. The tools that help — Language Reactor, Migaku, and the like — are browser extensions bolted onto Netflix or YouTube: great while you're inside the streaming tab, but they don't touch the video file on your drive, the podcast in your library, the PDF you're reading, or the ebook on your shelf.

The idea: give Langcap the whole source

Instead of studying alongside one streaming site, you hand Langcap the source itself — a video, an audio file, an ebook, a PDF, a web page — and it does the mining for you. It reads the whole thing, finds what's worth asking about, and turns it into a stream of bite-sized questions, each one linked back to the moment in the source it came from. No deck to build, no lesson to sit through. Answer one, and the next is ready.

Everything you add joins a library — a shelf of sources, each with its own cover — and every source can be studied two ways:

• Play — watch or listen it back with each word lighting up as it's spoken and a translation beneath. The comprehensible input you already like, now readable.
• Recall — drop into the question loop scoped to just that source, every prompt mined from its own lines, asked again until it sticks.

How it works — and why it stays on your machine

The part we care about most: it all runs locally. The speech and language models download once, then run on your own machine. Concretely:

• It transcribes on your GPU. Whisper turns audio and video into text on your hardware (a Vulkan-capable GPU); where a video already has subtitles, it uses them.
• It stitches every question to its source. Forced alignment maps the text back to exact timestamps, so each question carries the clip, line, or page it came from — not a detached flashcard.
• It schedules what you see. A spaced-repetition scheduler tracks each word in each direction and weighs what you missed, what's due, and what's just past your level to choose the next question for you.

No account, nothing uploaded, and after the one-time model download it works offline. Your files stay in your folders.

There's an optional on-device AI you can turn on to steer the loop in plain words — "only the kitchen-scene vocabulary," "just this show," "harder listening." It runs on your machine too, never the cloud. It's a steering shortcut, not the product: the questions themselves are chosen by a deterministic scheduler over the aligned source, not written by a model.

What's ready today

Langcap is a Windows founding beta, and it's honest about its edges:

• Seven languages have the full on-ramp — study from your own video, audio, or text: English, Spanish, French, German, Italian, Portuguese, and Chinese.
• Japanese and Korean study from audio today — drop in a podcast, a video, an audio file — with their synthesized voice for the text on-ramp coming next.
• Windows first (it needs a Vulkan-capable GPU); macOS later.

It's a young build, and the founding price reflects that.

Get it

Langcap is a one-time purchase — you own it, and it keeps working whether or not you ever pay again. It's $89 at launch; the founding beta is $59 (going to $89), with the first year of updates included and an optional $39/year after to keep getting new ones. Every purchase carries a 30-day money-back guarantee — if it's not right for you, every cent back.