Last updated: 7 October 2026
PikTalk is a voice-dictation keyboard for iPhone and iPad. This page describes exactly what it does with your data.
The short version
PikTalk has no accounts and no servers of its own. There is nothing to sign up for, and the developer receives no data from the app — no audio, no transcripts, no analytics, no crash reports, no identifiers. Your audio goes only to the speech-to-text provider you configure, and the transcript only to the polishing provider you configure. Everything else stays on your device.
What happens when you dictate
- Microphone. The PikTalk app records; the keyboard never records (iOS does not allow keyboard extensions to use the microphone). While a dictation session is armed the microphone input is open so recording can start instantly, which is why iOS shows its microphone indicator for the whole session. You can end a session at any time by long-pressing the keyboard’s mic key, and you can turn the armed background session off in the app’s settings (“Ready in the background”).
- Speech to text. The audio clip (16 kHz mono WAV) is uploaded over HTTPS to the transcription endpoint you configured — Groq by default, or any OpenAI-compatible endpoint you enter yourself. Nothing else is sent with it, and the clip is deleted from your device as soon as the request finishes.
- Polishing. The resulting transcript is sent over HTTPS to the chat provider you configured, which returns the cleaned-up text. If that step fails, the raw transcript is used instead.
- Insertion. The finished text is inserted into the text field you were already in, through the PikTalk keyboard extension.
What PikTalk never does
- It does not read, log or transmit your keystrokes.
- It does not contain analytics, advertising, tracking or crash-reporting SDKs, and it does not use an advertising identifier.
- The keyboard reads the surrounding text field for one purpose only: to check that the text it is about to replace is text PikTalk itself inserted, so a correction cannot overwrite something you typed. That text is never stored, logged or transmitted.
- It does not send your data to the developer, or to anyone other than the providers you configured.
Your API keys
PikTalk brings your own provider keys. They are stored in the iOS Keychain, in a keychain group shared only by the PikTalk app and its keyboard, and they are sent only to the provider they belong to. They never pass through the developer.
What is stored on your device
Settings, dictation history, learned corrections, the pending transcript and a diagnostic session log are stored locally, in the app’s container and a shared app-group container (so that the keyboard can insert the text the app produced). You can clear history, learned corrections and the session log in the app’s settings, and deleting the app removes all of it. Audio clips are not kept: each one is deleted once it has been transcribed.
The keyboard and Full Access
PikTalk’s keyboard needs Allow Full Access so that it can reach the network (to ask the app for status) and write to the shared container. It uses that access only for dictation. The keyboard talks to the app over a local connection on the device itself (127.0.0.1), authenticated with a token generated for your install; nothing is exposed to your network.
Third parties
Audio and transcripts are processed by the providers you chose, under their own privacy policies (for example Groq, DeepSeek, OpenAI, or whichever compatible endpoint you enter). The developer has no access to that processing. Beta builds distributed through TestFlight are also subject to Apple’s own terms and privacy policy.
Children
PikTalk is not directed at children under 13, and the developer collects no personal information from anyone.
Changes
If the app’s behaviour changes in a way that affects this policy, this page will be updated and the date at the top changed.
Contact
Questions about this policy: atom2ueki@gmail.com.