Doing

Doing is a desktop app for Mac that turns your voice into text instantly, right on your machine. Hold a hotkey, speak, and release — your words appear at the cursor, pasted into any app. There's no cloud upload, no account, and no monthly fee. You pay $49 once and own it. The app also lets you drag to capture screenshots while recording, attaching them to your transcript automatically. It's built for people who spend their days typing — especially developers working with AI coding tools like Claude Code, ChatGPT, or Cursor.

The transcription runs locally using NVIDIA's Parakeet model, which processes 60 seconds of audio in about 400 milliseconds. Doing includes a built-in benchmark tool so you can compare Parakeet against Apple's local engine or cloud services like OpenAI Whisper, Gemini, and AssemblyAI (using your own API keys). For post-processing, Skills let you apply prompts that clean up filler words, formalize tone, or convert speech into a code prompt — and they automatically switch based on which app you're pasting into. YOLO Mode makes it even faster by pressing Return automatically after each transcription, so your prompt runs without extra clicks. Audio ducking fades your music when you start recording and brings it back when you stop.

The target audience is anyone who types a lot, especially AI developers who issue hundreds of prompts daily. Instead of typing every prompt, you talk to your tools. Doing works anywhere you can type: terminal, browser, editor, or chat app. It's a one-time purchase from an independent developer, no strings attached. If you use AI coding assistants and want to move faster, this is a practical, no-nonsense tool that does one thing well.