Launching new features in OpenSS

OpenSS started as a fast way to capture what is on your Mac and ask a useful question without leaving the task. The newest work keeps that idea, but tightens the experience around it: less setup, more local processing and clearer recovery when a component is not ready.

Speak, review, then send

Voice input now uses Apple-native audio capture and on-device speech recognition in the macOS desktop app. You tap the microphone, speak, and see the transcript in the composer. A short correction window gives you time to stop or edit automatic submission.

Raw audio is not sent to OpenSS, OpenAI or Anthropic. The text follows the same route as a message you typed: OpenSS-hosted AI when you choose the hosted path, or the selected provider when you use your own key.

A local engine that can recover itself

The desktop interface and bundled engine now verify that they agree on an API version before chat and capture become available. If an older engine survived an app replacement, OpenSS can identify confirmed OpenSS-owned processes, replace the stale instance and retry with the bundled version.

The recovery interface has one clear action—Restart and retry—rather than asking people to choose between technical restart levels.

Hosted AI with clearer accounting

Free includes 200 AI units with Mini. Pro remains $15 per month with 500 weighted AI units and access to Mini, Luna, Terra, and Sol. Optional provider keys remain a Terminal CLI workflow for advanced users.

The Account page now focuses on units used, included units remaining and additional AI credits. It no longer needs to expose OpenSS internal model cost, which could be mistaken for a user charge.

Capture stays native

Chrome, Microsoft Word and PowerPoint remain the supported capture targets. Screen capture and OCR begin locally, then the relevant submitted context follows the hosted or BYOK route you selected. Voice and capture failures now appear as useful chat feedback rather than unexplained generic toasts.

Documentation for every surface

The new documentation separates the Apple Silicon desktop app, Terminal CLI and local browser GUI. That matters because they do not share the same account requirements: the browser GUI is served from 127.0.0.1 by openss gui, while the public website handles product information, accounts, billing and documentation.

What comes next

We are continuing to qualify the complete macOS path—from signed DMG to local engine, capture, voice and hosted authentication—before each public desktop release. Intel desktop packaging remains separate work and is not advertised as available.