Most voice apps trade your privacy for AI features. Offscript changes the game by bringing powerful transcription and summarization directly to your device. No accounts, no internet required, and zero data harvesting. • Secure: Industry-standard encryption keeps your recordings safe. • Intelligent: Use semantic search to find "that one idea" instantly. • Flexible: Organize with color-coded tags and export in multiple formats. Finally, a pro voice recorder that respects your data.

Hi Product Hunt! I’m the creator of Offscript, and I built it because I was tired of the "privacy tax" in the AI era. We’ve all seen amazing AI voice tools lately, but most of them come with a catch: they upload your most intimate thoughts, meeting minutes, and private brainstorms to a cloud server to be processed. As a developer and someone who values privacy, that never sat right with me. I wanted a professional-grade voice memo app that gave me the power of OpenAI’s Whisper and Semantic Search, but with the security of an air-gapped device. Why Offscript is different: 100% On-Device AI: Transcription and summarization happen on your phone's chip. No audio ever leaves your device. Search by Meaning: You can search for "that idea I had about the marketing plan" and our local AI finds it, even if you didn't use those exact words. No Accounts: No sign-up, no tracking, no data harvesting. Just hit record. Pro Features: Export to Markdown/PDF, biometric locking, and color-coded tags to keep your life organized. I’m sticking around all day to answer your questions about the tech stack, on-device AI, or our roadmap! I’d love to hear your feedback—especially from the privacy enthusiasts and "tools for thought" community. What’s one feature you’ve always wanted in a voice recorder but were too afraid to use the cloud for? Happy recording!

What model are you running on device for the semantic search, and how does it perform on older Android hardware with longer recordings? Nice work on keeping everything local!

@mateuszjacni Thank you! We're using a version of whisper for transcription and MiniLM for semantic search. So far so good with older devices. There were a few tricks involved to get that part "right" after some initial testing

What model are you running on device for the semantic search, and how does it perform on older Android hardware with longer recordings? Nice work on keeping everything local!

Offscript

Private AI voice memos that never leave your device.

Private AI voice memos that never leave your device.