CaptionNow 廣東話即時字幕・語音轉文字 vs Lugji

A real-time speech-to-text tool built around two things: accuracy and low latency. The developer is hard of hearing, and couldn't find a single app that handled Hong Kong Cantonese properly — the code-switching between Chinese and English, the technical terminology — while staying accurate and returning results instantly. So this one got built. Every decision in it comes from what actually holds up in daily use, not from a feature checklist. Features Live transcript — Text appears as you speak. No waiting for the sentence to finish. Captions for audio playing on your phone — Not just the microphone. Video calls, videos, and podcasts get transcribed too, which most comparable tools can't do. Floating captions — Picture-in-Picture keeps captions on screen while you use other apps. You pick the language; it doesn't guess — Mandarin (Taiwan / Mainland), Hong Kong Cantonese including Chinese-English code-switching, and English (US / UK / India / Hong Kong). Once you choose, it stays there — no drifting between languages mid-sentence. Offline recognition — Download a language pack on iOS 26 or later and it works without a network connection. Built for long reading sessions — No decorative clutter. Comfortable to watch through an entire lecture. Who it's for Deaf and hard-of-hearing users — Everyday communication support. See what's being said. Students — Live lecture notes, without losing the thread while you try to keep up. Professionals — Meeting minutes and interview transcripts, handled as they happen. Specifications This is transcription, not translation. What goes in is what comes out — with one exception: spoken Cantonese is transcribed into standard written Chinese, with English terms kept inline as spoken. Online mode uses a cloud recognition engine. Offline mode requires iOS 26 or later. Real examples of recognition errors are welcome — they're the most effective way to improve accuracy right now. Terms of Use (EULA): https://garnet-marimba-2dd.notion.site/Terms-of-Service-2d2f598ba98e80e5b065d198d8acf404?source=copy_link Privacy Policy: https://garnet-marimba-2dd.notion.site/Privacy-Policy-2d2f598ba98e80af99bae73136d3a3b9?source=copy_link

Let every conversation become knowledge you own. Lugji is an on-device voice assistant built for English, Cantonese & Mandarin meetings. Recording and transcription run entirely on your iPhone, then write into your iCloud or the Obsidian vault you connect — no account, nothing sent to a processing server. Lugji is serverless Your content and keys never reach Lugji. No backend, no API keys, no tracking. Privacy here isn't packaging — it's a structural fact. Core features • Meeting recording & live transcription — one-tap recording for long meetings and interviews; Cantonese, Mandarin and English transcribed live with Apple's built-in on-device engine. • AI summary (on-device · helper) — on Apple Intelligence–capable devices, Lugji can generate a meeting summary, processed entirely on your device: no Apple servers, no Private Cloud Compute, no cloud. (Requires iPhone 15 Pro or later; not available in mainland China.) • Quick Notes — a thought, a whiteboard photo, a voice snippet: captured instantly as standard Markdown. Apple Pencil sketches, image text recognition (OCR · Traditional Chinese + English), and voice-to-text. • Write into iCloud or Obsidian — every note is saved to your iCloud, or the local Obsidian vault you connect, with frontmatter and tags. Obsidian is your library; Lugji is the fastest way in. Why on-device? Cloud transcribers upload your audio to servers where it may be used to train AI; English-first models may be less suitable for Cantonese conversations; and they make you download 1–2 GB models and set up API keys. Lugji uses Apple's built-in engine, works in airplane mode, and is tuned for Cantonese meetings and everyday conversation. Device requirements: on-device transcription needs iPhone 12 / iPhone SE (3rd gen) / iPad (10th gen) / iPad Air (4th gen) / iPad mini (6th gen) / iPad Pro (11" 3rd gen, 12.9" 5th gen) or newer. AI Summary requires an Apple Intelligence-capable device. Recording itself works on every supported device. Pricing Start free (with usage limits). Plus (unlimited) and Ultra (advanced security: on-device AES-256-GCM encryption for locked notes & recordings, key only in your iCloud Keychain · enterprise) are planned for a future update. — Made in Hong Kong. Apple and Apple Intelligence are trademarks of Apple Inc. Obsidian is a trademark of Dynalist Inc.