Talking to AI is faster than typing to it
Why dictation is the best way to write prompts for ChatGPT, Claude and coding assistants — and a few habits that get you better answers.
The best prompts are long. They explain the situation, the constraint, what you already tried and what a good answer looks like. That's exactly the kind of writing people cut short when typing — and exactly what's easy to say out loud.
Why voice works so well for prompts
- You give more context. Saying three extra sentences costs seconds. Typing them feels like work, so we skip them.
- Conversational is fine. Models read "so basically what I'm trying to do is…" perfectly well.
- You stay in flow. Hold a key, think out loud, let go, read the answer.
A prompt you'd never bother typing
Write a Python function that removes duplicate rows from a list of dictionaries and keeps the first occurrence. The rows come from a CSV export, so the values are all strings, and two rows count as duplicates if the email field matches ignoring case. Keep the original order and add a short docstring.
That's 15 seconds of talking. With Sayburst it's on the screen the moment you let go — already capitalized and punctuated.
Habits that get better answers
- Lead with the goal, then the details: "I need X. Here's the situation."
- Say what you've tried. It saves a round trip.
- Name the format you want: "a table", "three options", "just the code".
- Use your vocabulary list for library names, product names and teammates, so they come through spelled right.
Asking without typing at all
Sayburst's Ask shortcut sends a spoken question to the writing provider you've set up — a local model through Ollama or LM Studio, or a cloud API with your own key — and shows the answer in a small panel. Nothing is inserted unless you click Insert.
Hold fn, think out loud, let go. Try Sayburst free for 7 days.
Download SayburstQuestions people ask
Can I dictate prompts into ChatGPT, Claude or Cursor?
Yes. A dictation app that pastes into the focused field works in any chat box or editor, including ChatGPT, Claude, Cursor and VS Code.
Should I clean up dictated prompts?
Lightly. Removing fillers helps, but models handle conversational phrasing well. Context matters more than polish.