开口兽 Help and Support

Help and support

Hangzhou Xulian Information Technology Co., Ltd. provides learning, word/sentence-query, account and independent-feedback support for 开口兽. Email: info@xunova.tech.

Mention 开口兽, your app version and OS version. Do not send passwords, Apple verification codes, session tokens or sensitive query content.

Privacy policy

Terms of service

Help and support

Contact us

• Email info@xunova.tech or open Feedback in app Settings. Feedback does not require a learning account and is separate from accounts and collections. Read replies in the original device’s conversation; no automatic email or push is sent. Access after reinstalling or changing devices is not guaranteed.

Email support

Word lookup and translation

• Type in the same input or tap Paste on its right. English-to-Chinese (en-zh) is the default: English words show dictionary entries and sentences show Chinese translations. Select Chinese-to-English (zh-en) for English translations of Chinese words or sentences with a short Chinese explanation. After explicit Microsoft Azure OpenAI permission, queries run automatically after an input pause without a permanent send button. Both translation directions and lookup share usage limits and require a connection and sign-in. AI output is not an authoritative dictionary.
• Withdraw cloud-processing permission in Settings to stop future queries. This does not disable local English playback of loaded text; previous third-party processing cannot be undone. Fetching or reloading scenarios separately requires sign-in and a connection. Only deliberately saved words or sentences enter a personal collection; ordinary sentence queries are not account history.

Loading scenarios and tracking check-ins

• Sign in and connect to load the scenario catalog and questions, then choose a category. Each scene has a stable unique ID and revision; material stays in current-run memory, not a bundled or downloaded offline bank. Initial loading, restarting and reloading require a connection. Retry reported loading errors rather than treating an empty screen as a completed download.
• Deliberately check in or undo according to your own experience. The account keeps completed scene IDs to count completed/untried material and retain earlier progress, not score ability. A read-only account snapshot does not contain offline question text. Signing out does not delete cloud progress; deleting the account removes its records.

The blue voice-input button

• Tap the blue button to request microphone and system speech-recognition permission. English-to-Chinese recognizes English; Chinese-to-English recognizes Chinese. Tap Stop or pause speaking. Final recognized text fills the input, then queries automatically according to your cloud-permission choice.
• Recognition is strictly on-device. Unsupported devices or languages report an error without cloud-recognition fallback; type or deliberately paste instead. Audio and transcripts stay only in memory and the app does not save or upload recordings. Cancelling, leaving the page or moving to the background stops capture. Local recognition does not make translation offline: Azure AI queries on final text still need separate permission, sign-in and a connection.

Accounts and deletion

• Use the same Sign in with Apple identity to restore an account. A name and avatar are optional and can be edited; an avatar can be deleted in Settings. Collections must sync online before cross-device restoration. Signing out does not delete them.
• Delete Account in Settings removes personal profiles, avatars, cloud collections, scenario check-in state and related sessions. Retry reported failures; signing out is not a substitute for deletion. Independent support conversations are not deleted with the account and need a separate request.

Playback, content and supported devices

• The complete Kokoro model ships with the app; its initial loading takes longer. Each playback performs fresh inference directly into memory without saving, caching or reusing generated audio or uploading speech text. Keeping the loaded model and voice in the current process is not audio caching.
• The learning interface is primarily Chinese, with original English/Chinese scenarios, blind listening, bilingual understanding and shadowing. Speech input is separate from English playback; there is no pronunciation scoring or claim of full Japanese or Korean interfaces. The first planned release is for iPhone and iPad on iOS/iPadOS 18.6 or later; public download is not available yet and the Mac app remains in development. Local recognition additionally depends on device and language support.