Lesson checklisttap to expand
01Free preview

Voice-to-AI: Whisper Flow setup + first prompt

Install Whisper Flow, assign a hotkey, and send your first voice-dictated prompt to an AI in under 15 minutes.

15m
# Voice-to-AI: Whisper Flow setup + first prompt In the next 15 minutes you're going to install a tool that turns your voice into text anywhere on your computer, assign it a hotkey, and send a real AI prompt without typing a single word. That's the whole workshop. One setup, one prompt, one habit seeded. ## Why voice matters for AI work Typing is slow for thinking. When you're brainstorming, exploring a problem, or trying to get a messy idea out of your head, the gap between your brain and your keyboard introduces friction that flattens the output. Speaking is different. People naturally talk faster than they type — usually 3–5x faster. More importantly, spoken language is less edited. You say the actual thing, not the polished version of the thing. That rawness is often exactly what produces better AI output, because you're giving the model more of your real thinking, not the cleaned-up version. Voice-to-AI isn't about accessibility. It's a quality and speed improvement for everyone. ## Install Whisper Flow Go to [wisprflow.ai](https://wisprflow.ai) and download the app for your operating system. Whisper Flow runs locally or uses a fast cloud transcription model — the free tier is enough to complete this workshop. Install it, launch it, and give it microphone access when prompted. If you're on Windows or prefer an alternative, Superwhisper and MacWhisper are solid alternatives with similar hotkey behavior. The workflow below applies to any of them. ## Assign a hotkey Open Whisper Flow preferences and look for the "Push to talk" or hotkey setting. Set it to something you can hit without looking — common choices: - **Right Option** (if you're on a Mac and rarely use it) - **Fn** (lower left on most keyboards) - **Caps Lock** (remapped — Whisper Flow can intercept it on most systems) Avoid hotkeys that conflict with browser shortcuts (Cmd+Shift combinations, F-keys). You want something that fires immediately when you're mid-thought. Test it once right now. Hold the key, say "testing one two three," release. Confirm the text appeared wherever your cursor was. ## Your first voice prompt Open a new chat at [claude.ai](https://claude.ai), [chatgpt.com](https://chatgpt.com), or wherever you normally work. Click into the message box so your cursor is there. Hold your hotkey and dictate this: > "I'm working on [describe the problem or project you're currently thinking about]. What are the three most important questions I should be asking before I go further?" Release the key. Read what appeared. Fix any transcription errors — there will usually be one or two on the first try, especially with proper nouns or technical terms. Then send it. That's a voice-to-AI prompt. You just did it. ## What to expect from the transcription Whisper-based transcription is good, not perfect. A few things to know: **Punctuation is guessed.** The model infers sentence breaks and punctuation from your speaking patterns. Speak in complete sentences and pause at natural breaks to get cleaner output. **Proper nouns trip it up.** Brand names, tool names, and people's names often transcribe phonetically rather than correctly. Scan your output before sending. **Speed matters less than clarity.** You don't need to speak slowly. But mumbling or trailing off at the end of sentences produces garbled transcriptions. Finish your sentences out loud. ## The habit to build The point isn't to dictate every message. It's to use voice when you're in ideation mode — when you're fuzzy on a problem, brainstorming options, or trying to think through something complicated. Type when you need precision. Speak when you need volume and speed. The next workshop in this series uses voice as the first step of a research pipeline. Voice-brainstorm into AI, extract themes, build a research prompt. Today you set up the tool. Next time you'll use it for real work. --- Now you can apply this to: - Any AI chat where you want to think out loud instead of type - Voice-drafting emails or Slack messages before refining them - Quickly capturing ideas when you're away from your desk but near a laptop