Last updated: August 2026
The short version: Sparkle does not collect, store, or transmit any of your data. Everything happens on your device. There is no personal data for us to process, so privacy laws like GDPR are satisfied by design.
None.
Sparkle is designed from the ground up to be completely private. Everything is processed on your device and nothing is collected, so there are no servers and no personal data on our side.
We do not collect:
Sparkle uses on-device machine learning (WhisperKit, or Apple's built-in speech recognition on iOS 26) to transcribe your voice to text. All processing happens entirely on your iPhone. Your voice data never leaves your device.
When you enable Sparkle's keyboard, iOS asks for "Full Access." This permission is required for the keyboard extension to communicate with the main Sparkle app for transcription. Despite the scary-sounding name, Sparkle:
The "Full Access" permission is only used so the keyboard can signal the main Sparkle app to start and stop recording, and receive the transcribed text back. The recording itself happens in the main app, because iOS does not allow keyboards to use the microphone. All of it stays on your device.
Sparkle stores your recent transcriptions (last 20) locally on your device only. If a recording could not be transcribed, its audio is kept alongside the history entry so you can play it back and retry. That audio is deleted the moment its entry is deleted. This is a convenience feature so you can recover a transcription if something goes wrong. This data:
Your data is yours to delete at any time: clear history in the app's settings, or simply uninstall the app. Nothing remains anywhere else.
Sparkle does not use any third-party analytics, advertising, or tracking services. We do not integrate with any cloud services.
If you choose to email us a problem report from Sparkle's settings, it is sent by you, from your own mail app, and you can read every part of it before you send it. The attached diagnostic log records technical events and numbers, such as when a recording started and how long a transcription took. It never contains what you dictated, and no audio is ever attached. We use it only to debug the problem you reported.
Sparkle downloads its speech-recognition models once, from WhisperKit's model repository (hosted on Hugging Face). On a model's first load it also fetches a small tokenizer configuration file from the same service. These are ordinary file downloads: the request contains no user data, and nothing you dictate is ever transmitted. After that, transcription works fully offline. If you use Apple's built-in speech recognition, iOS may download language files for a new language directly from Apple. That request is made by iOS, not Sparkle, and it also contains nothing you dictated.
Sparkle does not collect personal information from any users, including those under the age of 16. Since we collect no data whatsoever, Sparkle is safe for users of all ages.
If we ever update this privacy policy, we will post the changes here and update the "Last updated" date above.
Data Controller: Sparkle App, based in Greece, EU
If you have any questions about this privacy policy, you can reach us at hello@getsparkle.app