Capture and recognition
Voice, text, screens, and meetings are inputs to Talkie. An on-device speech model processes audio locally after download. Selecting a cloud transcription service changes that path: the service receives the audio it transcribes.
Sync between devices
Sync is separate from recognition. Moving a memo between devices uses the configured sync service, even when its transcription was produced locally. Review your sync settings alongside your model settings.
AI and agent handoff
A recap or agent task can use a different provider from speech recognition. When you invoke a remote provider, the transcript, prompt, or attachments supplied to it are sent to that provider. Your own API key selects an account; it does not make remote processing local.
Common questions
Does local-first mean every feature works offline?
No. Downloaded on-device speech models can run locally. Sync, cloud transcription, and remote AI or agent services need network access.
Are my API keys the same as on-device processing?
No. Using your own API key connects Talkie to your provider account. Requests still go to that provider. Select an on-device model when you want local speech recognition.
Where can I read the privacy policy?
The Privacy link below opens the published Talkie privacy notice. This page explains processing choices; the policy describes the wider privacy terms.