Privacy policy

Waffler keeps the microphone open while it runs, holding only the last half-second in memory so your first word is not clipped. Nothing is kept or sent unless you dictate: a recording can include up to half a second from before you hold down your hotkey, and runs until just after you let go, or in hands-free mode until you stop it. The audio is sent directly from your device to the speech-to-text provider you chose (Groq, or OpenAI as a backup) using your own API key. When OpenAI does the transcription, your custom vocabulary goes with the audio as a spelling hint. Groq does not receive it. The transcript is then sent to your clean-up provider to be tidied up, again under your own key. We never see any of it. Your transcription history is saved on your device only, in ~/.waffler-hosted/history.json on macOS, or %USERPROFILE%\.waffler-hosted\history.json on Windows.

Waffler keeps your last 10 recordings as audio files in ~/.waffler-hosted/debug_audio (%USERPROFILE%\.waffler-hosted\debug_audio on Windows). They are kept only so transcription faults can be diagnosed. Older ones are deleted automatically as new ones arrive, and you can delete the folder whenever you like. If transcription fails, for example because a VPN is blocking the provider, the recording is saved to ~/.waffler-hosted/unsent (%USERPROFILE%\.waffler-hosted\unsent on Windows) so your words are not lost. These are not deleted automatically. None of these files are uploaded anywhere.

Your API keys (Groq, OpenAI and the optional Cerebras key) are stored on your device only, in ~/.waffler-hosted/.env on macOS, or %USERPROFILE%\.waffler-hosted\.env on Windows. They are never transmitted to us. All API calls go directly from the Waffler app on your machine to the provider’s servers. We are not in the middle.

Finished text is placed on your clipboard so it can be pasted where you were typing. During first-run setup only, when you reach the API-key step, and each time you switch back to the setup window, Waffler checks the clipboard for an API key you have just copied, so you do not have to paste it by hand. Anything not shaped like an API key is discarded straight away: it is never shown, logged or stored.

A few seconds after it starts, Waffler asks GitHub (api.github.com) whether a newer release exists, and when you choose to update, the new version is downloaded from GitHub too. Apart from the AI providers you set up, these are the only network calls the app makes. None of them carry your audio, your dictations or your keys.

No Waffler account is required to download or use Waffler. Waffler is free and open source. There are no purchases, subscriptions or licence keys. We do not collect or store any personal data.

This website runs no analytics and no tracking scripts, and it sets no cookies.

• Audio recordings
• Transcription text
• Your API keys
• Usage patterns or analytics from the desktop app
• Personal information of any kind

Waffler connects directly to the providers you set up. Since you bring your own API key, how your data is handled depends on which provider you use:

• Groq

  Groq states that it does not use API input data to train its models. Check their current terms for how long API data is kept and where it is processed. Groq is Waffler’s default provider.

  How Groq handles API data

• OpenAI

  OpenAI states that data sent through its API is not used for model training by default, and, as of September 2026, that it may be kept for up to 30 days for abuse monitoring. Retention varies by plan and API tier, so check their current terms.

  How OpenAI handles API data

• Cerebras (optional, clean-up only)

  Settings has an optional Cerebras key. Cerebras does no speech-to-text, so if you add a key it only ever receives transcript text to tidy up, never your audio. Your vocabulary is not sent to the clean-up provider. Check their terms for how they handle API data.

  Cerebras privacy policy

If you have questions about this privacy policy, open an issue on GitHub.