Frequently asked questions

Real-time interpretation (Cantonese / English / Mandarin + 92 languages), offline audio file transcription, AI-generated meeting minutes + chapters + emotional insights, AI-drafted follow-up emails. PDF document translation that keeps the original layout (you get a translated-only and a bilingual side-by-side version). Plus a browser extension (bilingual web translation + YouTube/Netflix/Bilibili/iQiyi/Tencent Video subtitle translation + click-to-translate). In one line: turn your meetings from "listening" into "structured records you can revisit".

We focus on Cantonese-English-Mandarin business scenarios. Translation quality on Cantonese business terminology beats the big players (terms like Premis, private rack, 頭先 — which competitors get wrong — we handle correctly). Sub-second latency. Pricing is roughly 1/10 of iFlytek's simultaneous interpretation service.

In a quiet environment, Mandarin reaches 95%+ accuracy. Mixed Cantonese-English business speech reaches 90%+. Environmental noise, heavy accents, and fast speech can reduce accuracy. Using a microphone usually adds 5-10% accuracy.

Cantonese, Mandarin, English, Japanese, Korean, Thai, Indonesian, Vietnamese — 8 mainstream languages can be set to "Auto-detect". The detected language is then translated in real time to the target language.

Per-minute. 1 credit ≈ 1 minute of real-time translation. New users get 60 credits free on signup. Packs: Trial (¥10 / 60 min), Standard (¥100 / 600 min), Pro (¥300 / 6000 min — 70% off).

No. Only actual recording time is charged. Paused time does not count. Credits are deducted at meeting end based on cumulative active duration.

Upload an audio file and it's charged by actual audio duration, same per-minute rate as real-time. The page shows "Approx N credits" before upload so you can estimate.

1 credit per page, drawn from the same credit balance as interpretation. Charged only when the translation completes — failures cost nothing. The page count and estimated cost are shown right after upload.

No, credits are valid forever once purchased.

Alipay with instant credit. WeChat Pay coming soon.

Yes. For orders over ¥500, electronic invoices are supported. Email live@oriji.ai with your invoice info — sent to your inbox within 3 business days.

For unconsumed credits with valid disputes (e.g. system fault, double-charge), we will compensate the equivalent amount upon verification. Consumed credits are non-refundable. See Terms section 5.

Alibaba Cloud OSS. Only your account can access them. Auto-deleted after at most 7 days.

Permanently in your account. You can delete individual meetings anytime in the meeting detail page; this also clears the OSS audio.

No. We process data only to deliver the Service. De-identified statistics (with personal info removed) may be used for quality analysis. You can opt out via live@oriji.ai.

Yes. Please ensure you have legal authorization from all participants. Liability for unauthorized recording is yours. See Privacy Policy.

Yes. Email live@oriji.ai to request account deletion. All associated data is removed within 30 days.

No. Speech recognition and translation run on cloud models and require a stable network. Wi-Fi or 4G+ recommended.

Yes. Open the web app in WeChat / Safari / Chrome — mobile UI is fully supported. No App download required.

Yes. Click "Share" on the recording page to generate a link with optional password and expiry. Recipients don't need to register — just scan or open the link.

Speaker diarization is not supported in the current version. You can use "Mark this moment" to add notes at key points. Speaker separation is in development.

The frontend auto-reconnects (exponential backoff 1/2/4/8 sec, up to 4 retries). The backend ASR link also retries up to 5 times. Already-translated content is not lost.

YouTube, Netflix, Bilibili, iQiyi, Tencent Video. The extension auto-detects original subtitles and overlays bilingual translation in real time. Works on Chrome / Edge / Brave / any Chromium browser. Install from Chrome Web Store, or download .zip from the homepage and load manually.

Build your glossary (source → target word pairs) under "Glossary" and bind it when creating a meeting. Terms like Premis, Oracle, West Kowloon, BOC will follow your defined translations.

When creating a meeting, fill the "Quick terms" field (comma- or newline-separated). Recognition accuracy improves by 20%, latency unchanged. Terms expire when the meeting ends.

Yes. Fill "Meeting topic" with one sentence (e.g. "Technical discussion with client about vector DB solutions") and the AI biases recognition and translation accordingly.