The problem with typing AI prompts
Everyone who uses ChatGPT, Claude, or Gemini regularly runs into the same frustration: writing a good prompt takes effort. You need to be specific, provide context, specify the format, set constraints, and often iterate multiple times to get the output you want.
Prompt engineering — the art of writing effective AI prompts — has become its own skill. But most people don't have time to learn it, and even those who do still spend significant time writing and rewriting prompts.
What if you could just speak what you want, the way you'd explain it to a colleague — and have the AI figure out the rest?
That's the idea behind Voice-to-AI-Prompt.
How voice-to-AI-prompt works in AirTypes
AirTypes uses My Agent with a Prompt profile (or your own profile instructions). Whisper transcribes on your device; then your text is sent to your AI using your API key (BYOK) — directly to OpenAI, Anthropic via a compatible gateway, Ollama, OpenRouter, etc. AirTypes never stores your key or the request on our servers.
- You speak — start with your profile name in the first few words (e.g. "Prompt — I want…") or speak freely for a general assistant reply
- Whisper transcribes locally — raw text on your CPU, offline
- My Agent calls your AI — system prompt from your profile + your words (OpenAI-compatible chat request)
- Output is typed — the model's reply appears at your cursor in any app
Privacy split: transcription stays offline. My Agent needs network reachability to your provider. Configure endpoint, model, and key in Settings → My Agent (stored encrypted locally).
Real examples: what you say vs. what comes out
"So um I need like a cover letter for a software engineering job at a startup, I've been a backend developer for 5 years working mostly in Python and I want it to sound confident but not too formal."
"Write a confident but conversational cover letter for a software engineering position at a startup. The candidate has 5 years of backend development experience, primarily in Python. Tone: professional yet approachable, not overly formal. Length: 3 paragraphs."
"I need to explain recursion to a 10-year-old, make it simple with a story or analogy maybe about something they'd know like minecraft or lego"
"Explain the concept of recursion to a 10-year-old using a simple story or analogy from a game or toy they know (e.g., Minecraft or LEGO). Keep the explanation under 150 words and avoid technical jargon."
"Review this email I wrote and make it more direct, I think it's too long and kind of meanders before getting to the point"
"Review the following email and rewrite it to be more direct and concise. Lead with the main point, eliminate filler, and reduce the overall length by at least 30%. Preserve the original meaning and professional tone. [PASTE EMAIL HERE]"
Who benefits most from voice prompt engineering
Developers and engineers
Quickly dictate code review requests, architecture questions, debugging prompts, and documentation requests without interrupting your flow to type a precise prompt.
Writers and content creators
Describe the content you need conversationally — tone, length, style, audience — and get a ready-to-use prompt that produces better outputs than a vague typed request.
Business professionals
Draft email templates, meeting agendas, reports, and presentations by speaking your requirements naturally. The AI takes care of the prompt structure.
Non-technical users
If you've ever felt frustrated that AI tools don't understand what you want, voice-to-prompt bridges that gap. You don't need to know what a "system prompt" is — just speak what you want like you'd explain it to a smart assistant.
People with motor difficulties
For users who find typing difficult or painful, voice is the natural interface. Voice-to-AI-Prompt removes the barrier of needing to type precise instructions.
The future: voice as the new prompt interface
Prompt engineering is a skill that emerged because text interfaces require precise instructions. But speaking is fundamentally how humans communicate — we are far better at expressing nuance, context, and intent verbally than in text.
As AI becomes more integrated into daily work, the most natural interface for interacting with it is voice. Not because voice-to-text transcription is becoming better (though it is), but because the AI layer that sits between your speech and the final prompt is becoming capable enough to bridge the gap.
In the near future, you won't need to know how to write a prompt. You'll just speak what you want — and your chosen AI will shape the output.
AirTypes delivers that today with My Agent: your speech, your provider, your keys on your device — and Whisper still transcribes locally so your audio never hits our servers.
How to try voice → AI prompts in AirTypes
- Download AirTypes and complete free signup
- Open Settings → My Agent
- Add your API endpoint, model, and API key (stored encrypted on your device only)
- Create a profile named e.g. Prompt with instructions like: "Rewrite the user's speech as a clear, structured prompt for a large language model."
- Use the Agent hotkey, say "Prompt," then speak casually what you want
- The AI response is typed at your cursor — use it in ChatGPT, Claude, or any workflow
Whisper runs offline on your CPU. My Agent needs internet to reach your provider. Read the full My Agent walkthrough →