Privacy Policy

Auvi — Live Transcription & Translation

Last updated: 31 March 2026

Your microphone

Auvi needs your microphone to convert speech into text. Here is exactly what happens with your audio:

• It is processed entirely on your device by on-device models
• It is never recorded — Auvi does not create audio files. Ever. No sound recordings are saved to your device under any circumstances
• It is never sent to Creative Machines Limited or any server, cloud service, or third party
• It is never stored beyond your current session as audio (unless you enable Save Transcripts in Settings — which saves text only, not audio)
• It is discarded from memory the moment you stop transcription or close the app

To be clear: Auvi is a transcription app, not a recording app. It converts speech to text in real-time. No audio files are created, stored, or retained at any point. All speech recognition and translation runs locally on your device using on-device models. No audio or text data leaves your device at any time.

On-device processing

All speech recognition and translation runs locally on your device using on-device models. No audio or text data leaves your device at any time.

We do not train AI on your data

Creative Machines does not use any audio, transcription text, voice profile, translation, or any other data generated by Auvi to train, fine-tune, or improve artificial intelligence or machine learning models. We have no mechanism to access, retrieve, or collect such data from your device.

Ephemeral vs. persistent — you choose

• Default (ephemeral): Transcribed text lives in volatile memory only. Stop transcription or close the app, and it is permanently gone. No file, no log, no record.
• Persistent (opt-in): If you turn on Save Transcripts in Settings, text only is stored locally on your device. No audio is saved. Saved transcripts are never sent to Creative Machines or any third party.

Optional local storage

These features are off by default. They only activate if you explicitly turn them on:

• Save Transcripts — stores text locally on your device, excluded from iCloud backup
• Allow Copy — lets you copy transcription text to the clipboard and take screenshots
• Allow Export — lets you share or export saved transcripts

By default, Auvi prevents screenshots and screen recordings from capturing transcript content. This means the live captions on your screen will appear blank in any screenshot or screen recording. This protection is only lifted when you have purchased Plus and enabled Allow Copy — ensuring that transcript text cannot be inadvertently captured or shared without your explicit opt-in.

If you enable saving, you can delete individual transcripts any time or set automatic deletion (7, 30, or 90 days).

Vocabulary packs

Auvi includes optional domain vocabulary packs (e.g. Medical, Legal, Technology) that improve transcription accuracy for specialist terminology. These packs are static data files bundled with the app — they are not downloaded, personalised, or updated based on your usage. No vocabulary data is sent to any server. Enabling a vocabulary pack simply provides additional context to the on-device speech recognition model.

Voice profiles and biometric data

If you create a voice profile (optional), the voice embedding is:

• Stored only on your device
• Never sent to any server
• Used solely to distinguish your speech from other speakers
• Deletable at any time. When you delete a voice profile (Settings > Accessibility > Voice Profile > Delete), the embedding data is permanently and securely erased from your device. Creative Machines has no copy to retain or destroy, as the data never leaves your device.

Biometric data notice (BIPA): Voice embeddings may constitute biometric data under certain laws, including the Illinois Biometric Information Privacy Act (BIPA). Creative Machines never collects, possesses, receives, or otherwise obtains biometric data from users. Your voice embedding is created and stored exclusively on your device and is never transmitted to us or any third party. You may delete it at any time.

Retention schedule: Voice embeddings are retained on your device for as long as you choose to keep them. There is no automatic expiry. Embeddings are destroyed when you delete the voice profile, reset the app, or uninstall the app. The destruction method is secure deletion from local device storage.

Consent: Before creating a voice profile, the App requires your informed written consent by typing your initials, confirming you understand that a biometric identifier will be created and stored on your device. A local consent record (timestamp, initials, and device identifier) is stored on your device and is accessible in Settings > Voice Profile Consent History.

Safety features

Auvi includes optional safety features — sound awareness, safety keyword alerts, and scam detection (phone and in-person) — that analyse audio and transcribed text to alert you to potential hazards or suspicious conversation patterns. All analysis happens entirely on your device:

• No alert data, detection results, or conversation content is sent to Creative Machines or any third party
• No record of alerts is stored beyond your current session (unless Save Transcripts is enabled, in which case the transcript text — not the alert metadata — is stored locally)
• You can enable or disable these features at any time in Settings

Location tagging

If you enable location tagging (off by default), Auvi captures a single GPS coordinate when you save a conversation. This location is:

• Stored only on your device alongside the saved transcript
• Never sent to any server or third party
• Used solely to tag where a conversation took place
• Deletable at any time by deleting the saved conversation

Location data is only captured once per save — Auvi does not continuously track your location. This feature is opt-in and can be disabled in Settings. Corporate IT administrators can permanently disable it via MDM.

Apple Watch companion

If you use the Apple Watch companion app, live caption text is streamed from your iPhone to your Watch over Bluetooth via WatchConnectivity. This data:

• Travels directly between your paired devices over Bluetooth — never via any server
• Is not stored on the Watch beyond the current viewing session
• Is used solely to display captions on your wrist

Be aware that captions displayed on your Watch screen are visible to anyone who can see your wrist.

On-device language model

Auvi optionally downloads a small language model (Qwen3-1.7B, ~1.1 GB) to enhance conversation summaries, action item extraction, and title generation. This model:

• Runs entirely on your device — no text is sent to any server
• Is downloaded once from a public model repository (Hugging Face) and stored locally
• Processes only the transcript text of the conversation being saved
• Can be deleted at any time by removing the downloaded model in Settings

On newer versions of iOS (26+), the built-in Apple Foundation Models are used instead, requiring no additional download.

External display and lock screen

Auvi can display captions on an external screen (via AirPlay or HDMI) and show the most recent transcript line on your lock screen via Live Activities. These features are designed for accessibility — for example, displaying large captions on a TV during a group conversation.

Be aware that captions shown on an external display or lock screen are visible to anyone who can see that screen. No transcript data is sent to any server — external display output goes directly from your device to the connected screen.

Text-to-Speech

If you enable Translation, Auvi can speak translated text aloud using on-device text-to-speech. No audio is recorded or stored, and no data is sent to any server. The spoken audio is generated entirely on your device.

Model download

The only network request Auvi makes is a one-time download of on-device speech recognition models when you first set up the app. On newer versions of iOS, the required models may already be present on your device. After any initial download, the app makes no further network requests.

Translation

If you purchase the Translation add-on, translations are performed entirely on your device. No text is sent to external servers.

Managed devices (MDM)

For corporate deployments, IT administrators can enforce privacy settings via Managed App Configuration. When MDM keys are present, certain features (saving, copying, exporting) can be permanently disabled. The app does not communicate with any MDM server — it only reads the local configuration profile.

Data Controller

Creative Machines is not a Data Controller or Data Processor under GDPR (or equivalent laws) for any transcription data generated by the app. We never access, receive, process, or store any data generated on your device.

If you enable Save Transcripts or Allow Export, you are the Data Controller for that data. You are solely responsible for compliance with applicable data protection laws.

GDPR

Because Auvi does not collect, process, or store any personal data on any server:

• There is no personal data to consent to processing
• There is no data to request access to or deletion of
• There are no data transfers to third countries
• There are no data processing agreements required

The app is fully compliant with the General Data Protection Regulation (GDPR).

UK GDPR

Auvi is equally compliant with the UK General Data Protection Regulation (UK GDPR) as retained in UK law after Brexit. The same principles apply — no personal data is collected, processed, or stored on any server.

Brazil LGPD

Auvi complies with Brazil's Lei Geral de Proteção de Dados (LGPD). No personal data is collected, processed, or transferred. All processing occurs locally on your device.

CCPA

Auvi does not sell, share, or disclose any personal information. No data is collected from California consumers or any other users.

Hong Kong PDPO

Auvi is published by Creative Machines Limited, incorporated in Hong Kong SAR. As we do not collect personal data, the Personal Data (Privacy) Ordinance (Cap. 486) (“PDPO”) does not apply to our data handling practices. However, as a user of Auvi, you are the Data User (equivalent to Data Controller) responsible for ensuring your use of the App complies with the PDPO, particularly when transcribing personal data of others.

Children's privacy

Auvi does not collect any data from anyone, including children under 13. The app is rated 4+ and is suitable for all ages.

Changes to this policy

If we update this privacy policy, the updated version will be posted at this URL. Since we collect no data, changes would only reflect new features or clarifications.

Contact