Your privacy is our top priority. This policy explains how we handle your information.
1. About Us
YuSu (雨粟) Input Method is developed and operated by a non-profit organization based in New York, USA, built on open-source content. We are committed to providing a clean, secure, and trustworthy input method solution.
2. Keyboard Typing Input: Processed Locally, Not Collected
For your keyboard typing input, we never spy on your privacy:
Local Processing: Keyboard typing input is processed locally on your device, not uploaded to servers
No Tracking: We do not track individual users or usage behavior
No Accounts: Basic features do not require registration or personal information
No Monitoring: We do not monitor or collect your typing content
Exception: Audio is uploaded only when you select the online voice engine; local voice engines do not upload audio. See Section 3.
3. Voice Input Feature (Optional)
When you use the voice input feature:
Local Voice Engines: When you select the local SenseVoice or Qwen3 engine, recording and transcription occur on your device and the audio is not uploaded. On iOS, selecting the local SenseVoice engine requires a one-time download of a speech model of about 470 MB on first use; the model is stored only on your device, and the download uses mobile data and storage space.
Online nova-3: When you actively select online nova-3, the recording is uploaded to YuSu over HTTPS and forwarded to Deepgram's nova-3 service for transcription. Deepgram receives that audio as the speech-processing provider.
No Storage on Our Server: The audio is used only for that single transcription and is deleted immediately after processing. We do not store any recorded audio on our server, do not use it for model training, and do not link it to your identity.
Encrypted in Transit: Audio is transmitted over HTTPS throughout.
Only When You Actively Use It: Audio is uploaded only when you actively trigger voice input and record; no recording means no upload.
Polishing and Output Language Conversion: If you enable online polishing or select an output language that requires online translation, the transcribed text is sent to YuSu for processing. The original audio is not uploaded for this step.
In short
Local engines (SenseVoice, Qwen3): recording and transcription stay on your device. The audio is not uploaded.
Online nova-3 only: the audio goes over HTTPS to the YuSu server, which forwards it to Deepgram nova-3. It is deleted right after processing and not stored.
First use of local SenseVoice on iOS: the speech model (about 470 MB) is downloaded once from the YuSu server (/voice/models) and stored only on your device. Online nova-3 needs no download.
Polishing or translation: only the transcribed text is sent to the YuSu server (/voice/polish, /translate), never the audio.
4. Cloud Input Feature (Optional)
When you choose to enable the cloud input feature:
De-identification: Your input content is de-identified before processing
Anonymization: Processed content is converted to public anonymous information
Used Only for Hot Word Queries: Anonymous information is used solely for network hot word queries to improve input accuracy
No Personal Association: Query data cannot be linked to your personal identity
Can Be Disabled Anytime: You can disable the cloud input feature in settings at any time
Cloud context prediction (iOS, optional, off by default): When enabled, the keyboard connects to the network as needed (this requires "Allow Full Access") and periodically downloads anonymous context word pairs aggregated by the server, which are used on your device for candidate prediction. This feature only downloads and never uploads: nothing you type on iOS is uploaded. You can turn it off in settings at any time.
5. Cloud Sync Feature (Optional)
When you choose to use the KeyLockr cloud sync feature:
End-to-End Encryption: Your configuration data is end-to-end encrypted
Settings Only: Only syncs input method settings and personal dictionaries, not input content
KeyLockr Account: Requires a KeyLockr account for authentication
You Control Your Data: You can delete cloud-synced data at any time
6. Data Security
We take the following measures to protect your data security:
Industry-standard security measures
All network transmissions use encrypted connections (HTTPS)
Regular security audits and updates
We do not sell or share any data with third parties
7. Updates and Feedback
To provide updates and improve services, the app may:
Version Check: Periodically check for available updates
Feedback Submission: Collect information you provide when submitting feedback
Crash Reports: Generate anonymous diagnostic reports if the app crashes (requires your consent)
8. Dictionary Downloads
When you download dictionaries, we do not record any download statistics or personally identifiable information.
9. Children's Privacy
YuSu (雨粟) Input Method is suitable for users of all ages. We do not knowingly collect personal information from children under 13.
10. Your Rights
You have the right to:
Know how we process your data
Disable cloud features at any time
Delete cloud-synced configuration data
Export your personal dictionaries and settings
Completely uninstall the app and all its data
11. Third-Party Services
YuSu (雨粟) Input Method may include links to or integration with the following third-party services:
KeyLockr: Optional cloud sync service with its own privacy policy
Deepgram: Processes the recording only when you select the online nova-3 voice engine
Open Source Components: Built on open-source projects, following respective open-source licenses
Local voice engines do not rely on third-party speech services. Audio processed by online nova-3 is subject to Deepgram's service and privacy terms. We are not responsible for third-party privacy practices.
12. Policy Changes
We may update this privacy policy from time to time. For significant changes, we will:
Update the "Last Updated" date
Notify users within the app
Publish the updated policy on our website
13. Contact Us
If you have any questions or concerns about this privacy policy, please contact us:
infosafex.cloud
SafeX Inc. New York, USA
Important Notice: By using YuSu (雨粟) Input Method, you agree to this privacy policy. If you do not agree with any part of this policy, please do not use this software.