voice-matching

Voice Matching

Voice is the consistent identity — vocabulary, rhythm, worldview, humor type. Tone shifts with context (celebratory, serious, casual). You can write in someone's voice while varying tone. Get the voice right first.

Voice Decomposition

When given examples of a creator's content, extract these dimensions:

Vocabulary — what words do they reach for? What would they never say? How formal or casual? Do they use slang, and which kind? Do they curse? Do they use filler words deliberately?

Sentence rhythm — short and punchy? Long and flowing? Fragments as emphasis? How do they use line breaks? Do they build momentum or stay staccato?

Opening patterns — how do they start? Direct statement? Question? Story drop? Address the audience? Jump straight into the point?

Humor style — dry/deadpan, self-deprecating, observational, absurdist, or none? How often does humor appear? Is it the main mode or occasional seasoning?

CTA patterns — how do they ask for engagement? "Follow for more" or subtle? Do they even do CTAs? How do they end posts?

Emotional register — enthusiastic, chill, intense, vulnerable, authoritative, conspiratorial? What's their default energy?

Examples Over Descriptions

"I'm funny and relatable" tells you almost nothing. Five example captions tell you everything.

Always prefer analyzing actual content over self-descriptions. If you only have a self-description to work from, flag that the voice match will be approximate and ask for examples if possible.

When analyzing examples, quote specific phrases that capture the voice. "Notice how they always start with a one-word fragment" is more useful than "they have a punchy style."

Authenticity Markers

The small patterns that make content feel like a specific person, not just a generic voice:

• Characteristic phrases they return to
• Specific emoji they always use (or never use)
• How they address their audience (name for the community? "you guys"? never address directly?)
• Topics they always come back to, even tangentially
• Structural tics — always end with a question? Always use numbered lists? Always include a callback?

These markers are what separate "sounds like a person" from "sounds like THAT person."

Anti-patterns

Generic influencer voice — "Hey guys! So excited to share..." This is nobody's real voice. It's a template. If the content you're generating could be attributed to any creator, the voice matching has failed.

Corporate casual — reads like it was approved by a marketing team. Technically warm but substantively empty. "We're thrilled to announce..." energy.

Over-optimization — stripping personality to maximize engagement metrics. The quirks and rough edges ARE the voice. Smoothing them out makes content forgettable.

Voice drift — starting in the creator's voice but gradually reverting to default AI writing patterns. The longer the piece, the more this happens. Re-anchor to the original examples regularly. If you notice yourself drifting, stop and re-read the examples before continuing.