Shall We Talk turns speech into clean text on iPhone and Mac. Dictate into any app with a voice keyboard that types at your cursor, Shortcuts, or a Mac hotkey, and record meetings into speaker-labeled transcripts and structured summaries. Cleanup is a proofreader, not a rewriter: it removes fillers and misheard words but keeps your wording, tone and Chinese-English code-switching. Open source (GPL-3.0); bring your own speech and LLM API keys. Advanced from build 200 to 243 with GPT-6 Astra.

What became possible in your product with Astra that was not practical before?

Astra made the long, unglamorous middle of building a native product practical: carrying one idea through two operating systems, asynchronous audio, keyboard delivery, persistence and regression tests. Shall We Talk already existed before Astra. Its last committed iOS baseline before the September 3 release was build 200; by September 14 it reached 243. I built mainly in Codex, with earlier models establishing the foundation and Astra leading the later implementation, supported by other agents for review and second opinions. That chapter improved Chinese keyboard input, safer text handoff, recording lifecycle stability, and start/stop meeting shortcuts across an evolving Swift codebase. The leap was continuity: turning real usage failures into connected fixes, tests and release work while keeping the human in charge of product decisions. This is an existing product substantially advanced with Astra, not a claim that every line or numbered build was created by Astra.

Shall We Talk grew from everyday friction with voice input: recognizing Chinese words, moving dictated text into the right place, and switching between short notes and meeting recordings. The project brings these workflows into native iPhone and Mac apps. This launch shares the development source and its setup instructions. On iOS, dictation through another keyboard uses a clipboard-and-paste workflow; the meeting shortcut opens the app. I welcome feedback on the workflow and Chinese-language input experience.

What became possible in your product with Astra that was not practical before?

Shall We Talk

Voice dictation that saves transcriptions on iPhone and Mac

Voice dictation that saves transcriptions on iPhone and Mac