Voice-First Context Prompting
Speak prompts aloud to supply richer context with less friction
- Difficulty
- Starter
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 97%
Voice-First Context Prompting replaces terse typed instructions with spoken explanations. When typing, users often shorten thoughts, edit as they compose, and omit contextual details. Speaking allows them to describe the product, audience, goal, source, constraints, and uncertainty in a more natural flow. A transcription tool converts that explanation into a detailed prompt that gives the model more material for pattern matching and decision-making. The user then continues conversationally, reviewing results and adding context rather than attempting to engineer one perfect written instruction. The mechanism is especially useful for complex creative work, where tacit knowledge and qualifications materially affect quality, but it still depends on the user's editorial judgment to choose and refine the result.
Origin
Barbara Jovanovic reported that her AI outputs improved substantially after she stopped typing most prompts and began speaking them through Super Whisper. Extracted from Marketing Against The Grain.
Core principles
- 01Models perform better when they receive abundant relevant context
- 02Speaking naturally reveals details omitted from concise typed prompts
- 03Prompt quality improves when expression requires less editing effort
- 04Human judgment still directs the model after context is supplied
How to run it
- 1
Define the outcome
Begin by saying what you need the model to produce or decide.
Pro tip Describe the practical use of the output, not only its format.
Watch out A long spoken prompt without a clear objective can still be confusing.
- 2
Narrate the context
Explain the audience, product, situation, source material, and relevant history in natural language.
Pro tip Mention details that feel obvious to you because they are not obvious to the model.
Watch out More context helps only when it remains relevant to the task.
- 3
State constraints and preferences
Describe what to emphasize, avoid, verify, or preserve.
Pro tip Include examples of good and bad outcomes when available.
Watch out Do not assume tone or quality standards are implicit.
- 4
Transcribe and submit
Use speech-to-text to turn the explanation into a prompt and send it to the model.
Pro tip Correct transcription errors in names, figures, or specialist terminology.
Watch out Uncorrected transcription errors can distort important constraints.
- 5
Continue the conversation
Review the response aloud and supply corrections or additional context in subsequent turns.
Pro tip Treat prompting as an iterative discussion rather than a single command.
Watch out Do not accept a fluent result without evaluating its substance.
In the wild
An editor verbally explains a founder's product, desired audience, current campaign, and the type of insight needed. AI searches the supplied transcript with that richer context and returns candidate talking points for review.
→ The candidates are more relevant than those produced from a short typed request.
Common mistakes
Speaking without an objective
Verbal detail becomes noise when the desired decision or deliverable is not stated clearly.
Trusting raw transcription
Speech recognition can corrupt proper nouns, product details, or numerical constraints that affect the result.
Replacing judgment with context
A richer prompt improves model performance but does not decide whether the output is worth using.
Is it for you?
Best for
People explaining nuanced content, workflows, audiences, products, or desired outcomes to an AI assistant.
Not ideal for
Sensitive conversations, noisy environments, or precise commands that are clearer when written.
From the transcript
“One thing that has really helped me is I very rarely type into Chat GPT. I only use Super Whisper, so I use just my…”
“And the results for whatever reason are infinitely better because I add more context.”
“Chat GPT performs much better when you give it as much context as possible.”
From the episode
I Run a 6-Figure Agency with ZERO Employees (AI Only)
Barbara Jovanovic