meta: title: Practice — べんきょうBox description: Take a podcast, a clip, or a recording you already have, and turn it into a conversation you can study line by line. Or build one yourself. url: /practice format: yaml content: - type: page-header body: Take a podcast, a clip, or a recording you already have, and turn it into a conversation you can study. Or build one yourself, line by line. heading: Real audio, broken down line by line eyebrow: 03 · PRACTICE accent: practice - type: walkthrough body: Three taps from a podcast on your phone to a conversation you can shadow. heading: Real human Japanese, sliced into a conversation label: Audio Clipper steps: - kind: paragraph body: Open Audio Clipper from the home screen. Load an MP3 from your device, a podcast you've downloaded, a clip from a video, a voice note. Select the range you want to extract. - kind: screenshot slot: label: SCREENSHOT SLIVER title: Waveform with range selection meta: '4:1' aspect: 4/1 image: waveform-range.png - kind: paragraph body: The app splits the audio into segments. You assign each one to P1 (Person 1), P2 (Person 2), or none. Each turn is transcribed locally on your device. - kind: screenshot slot: label: SCREENSHOT SLIVER title: Segment list with P1/P2 assignment meta: '4:1' aspect: 4/1 image: p1-p2-assign.png - kind: paragraph body: This is the part most learning tools don't have. Real human Japanese has the rhythm, the half-swallowed particles, the natural sentence structures, the way a sentence actually ends. You can shadow against it, repeating after the speaker, mimicking their intonation and pace, until your voice catches their pattern. That practice is hard to replicate any other way. - kind: screenshot slot: label: SCREENSHOT SLOT title: Audio Clipper, full view meta: 4:5 portrait aspect: 4/5 image: audio-clipper.png - type: card-grid body: 'Two ways in: type each line, or record each line.' heading: Two ways in, no audio file needed label: Or build a conversation yourself intro: If you've got a dialogue from a textbook, a script you've written, or a real exchange you want to study, you can build a conversation directly. No audio file, no clipper. columns: 2 variant: surface cards: - label: Type each line title: Add lines as text body: Add each line of dialogue with its speaker. Lines stay as text until you ask for anything more, so creating a conversation this way is a no-network operation by default. - label: Speak each line title: Or record each line body: |- Record each turn yourself. The app transcribes it locally and stores the audio with the line, so you've got both from the start. Useful for working through a textbook dialogue out loud, or capturing an exchange from a video, a teacher, or a conversation partner. - type: strip body: Auto-play through the whole exchange, or step through line by line. heading: Two ways to play it back label: Playback orientation: text-right paragraphs: - Play the whole conversation through automatically, like you're hearing it live. Or step through line by line, at your own pace. - Auto-play is for listening practice, getting your ear used to the rhythm of a real exchange. Line by line is for shadowing, breakdown work, or just taking your time over something difficult. screenshot: label: SCREENSHOT SLOT title: Conversation playback meta: Auto-play and line-by-line · 4:5 aspect: 4/5 image: conversation-playback.png - type: card-grid body: 'Every line, the same on-demand tools: translation, audio, voice quality.' heading: What every line gets, on demand label: Per line intro: Whether the conversation came from the audio clipper or you built it by hand, every line has the same on-demand tools. Nothing happens automatically. Translation costs money, audio generation costs money, and neither runs without you asking. columns: 3 variant: surface cards: - label: Translation title: Tap to translate, line by line body: |- Tap a line to translate it on demand, using your provider. The translation includes the same word-by-word breakdown you get from Translate, with readings, dictionary forms, and tap-through to entries. Lines stay untranslated until you ask, so the page does not quietly hit the network in the background. - label: Audio title: Original where it exists, generated when you want it body: |- If the line came from the audio clipper or a recording, that's the audio you get, real, human, what you started with. For typed lines, you can generate audio on demand. Like translation, this only happens when you ask. - label: Voice quality title: Cascades from your settings body: |- Generated audio uses your TTS configuration. Connect a third-party voice for the highest quality, or rely on the OS voice. The OS voice can be improved a little by downloading higher-quality voices in your iOS settings. - type: card-grid body: |- Audio clipper transcription is local. Recordings you make are stored on your device. Conversations themselves, with their text and audio, are stored on your device. The network is only touched when you ask for it: a translation on a specific line, or generated audio with a third-party voice. Each one is an explicit tap, not a background call. heading: 'Default state: nothing leaves your device' label: What stays local columns: 1 variant: sub cards: [] - type: faq body: Common questions about the audio clipper, audio handling, and shadowing. heading: Quick answers items: - question: What audio formats can the audio clipper load? answer: 'MP3 files from your device. Pick anything you''ve already got: a podcast you''ve downloaded, a clip from a recording app, audio you''ve extracted from a video.' - question: Does the audio leave my device? answer: No. Audio you load into the clipper is processed and transcribed locally. Same for recordings you make yourself. Audio doesn't leave the device unless you explicitly ask for translation or generated audio on a specific line. - question: What's shadowing? answer: A practice technique where you repeat after a native speaker, ideally as they're still speaking, mimicking their rhythm, intonation, and pace. The audio clipper makes shadowing easier because the original audio is right there, paired with the transcription. - question: Can I import existing scripted dialogues? answer: Not at the moment. You can build a conversation by typing or recording each line. If a text-import path would be useful, get in touch. - type: prose body: More on the [FAQ page →](/faq). - type: card-grid body: 'Coming soon to iOS and macOS. Up next: study.' columns: 2 variant: sub cards: - label: Get it title: Notify me when it ships → body: Coming soon to iOS and macOS. Android later. href: '#notify' - label: Next title: Study → body: Two small games for when you've got five minutes. href: /study tile: dictionary schema: '@context': https://schema.org '@type': FAQPage name: Practice — べんきょうBox description: Take a podcast, a clip, or a recording you already have, and turn it into a conversation you can study line by line. Or build one yourself. url: https://getbenkyobox.com/practice mainEntity: - '@type': Question name: What audio formats can the audio clipper load? acceptedAnswer: '@type': Answer text: 'MP3 files from your device. Pick anything you''ve already got: a podcast you''ve downloaded, a clip from a recording app, audio you''ve extracted from a video.' - '@type': Question name: Does the audio leave my device? acceptedAnswer: '@type': Answer text: No. Audio you load into the clipper is processed and transcribed locally. Same for recordings you make yourself. Audio doesn't leave the device unless you explicitly ask for translation or generated audio on a specific line. - '@type': Question name: What's shadowing? acceptedAnswer: '@type': Answer text: A practice technique where you repeat after a native speaker, ideally as they're still speaking, mimicking their rhythm, intonation, and pace. The audio clipper makes shadowing easier because the original audio is right there, paired with the transcription. - '@type': Question name: Can I import existing scripted dialogues? acceptedAnswer: '@type': Answer text: Not at the moment. You can build a conversation by typing or recording each line. If a text-import path would be useful, get in touch.