Pat Simmons asks Claude Code to design a beginner-friendly macOS push-to-talk application. The resulting Swift app captures microphone audio, shows a small recording display, transcribes speech locallyAutomatic speech recognition converts spoken language into text or structured linguistic output using signal processing and machine learning. and inserts the text at the active cursor after the required accessibility and microphone permissions are granted.
Pat Simmons compares Apple's built-in Speech Transcriber with Nvidia's Parakeet modelModel selection chooses the AI model whose capability, quality, cost, speed, safety, and operating constraints best fit a task.. Both produce rapid local resultsLow-latency AI inference produces a model result quickly enough for an interactive or time-sensitive application., with Parakeet slightly faster in the demonstration, while the native Apple option offers the simplest setup for macOS users.
Pat Simmons then adds a desktop interface with transcription history and a custom dictionary that corrects recurring terms. The finished app keeps speech on the user's machineData privacy governs how personal, confidential, or sensitive information is collected, used, shared, retained, and protected in AI systems. and can be adapted without depending on a subscription dictation service.
Watch on YouTube
