# Voiceprint profile: [NAME]

Built [DATE]. Last re-measured [DATE]. Samples on file: [N].
Fill what you know, leave the rest empty. An honest gap beats a guess.

---

## Layer 1, Identity

- In one sentence, what I am for:
- Who I am talking to:
- What makes a reader say "that is definitely them":
- What I refuse to be:

## Layer 2, Beliefs

Positions I hold and defend in writing. Not private facts about me.

- The take I repeat that people argue with:
- Whose side I am on when the obvious two sides fight:
- Overrated in my field:
- Underrated in my field:
- What I will not pretend to believe:

## Layer 3, Decision patterns

The strongest recognition layer. What I do with information, not what words I use.

- What I deliberately leave out that others include:
- What I never write about:
- What I always soften:
- What I never soften:
- What makes me go blunt:
- What makes me go dry:
- When I let emotion in:
- Where I refuse to be helpful:

**My argument shape** (the order I build things in, from the samples):

- Default shape:
- Second shape, if any:
- Shape I never use:

## Layer 4, Writing patterns (measured, not guessed)

Numbers from [N] samples, [TOTAL] words, measured [DATE]. A fingerprint for
comparison. Never a quota to write toward.

| Measure | Value |
|---|---|
| Sentence length: shortest / median / longest | |
| Sentences under 6 words | % |
| Sentences per paragraph, median | |
| One-line paragraphs used | yes / no |
| Commas per sentence, median | |
| Colons per 1000 words | |
| Semicolons per 1000 words | |
| Parentheses per 1000 words | |
| Dashes per 1000 words | |
| Question marks per 1000 words | |
| Exclamation marks per 1000 words | |
| Rhetorical questions per 1000 words | |
| First person per 1000 words | |
| Average word length, characters | |
| Words over 8 characters | % |
| Share of text that is bulleted | % |
| Contractions used vs written out | % |
| **TTR** (note the word count too) | |
| **MATTR**, 100-word window | |
| **MTLD**, 0.72 threshold | |

- Connectors I use:
- Connectors that never appear in my writing:
- How I open a paragraph (the recurring first words):
- How I close:
- Characteristic irregularities (self-corrections, afterthoughts, asides):

## Layer 5, Surface style

- Signature words and phrases:
- Words I refuse to write:
- Casing and formatting habits:
- Slang level:
- Sign-offs:
- German only, my Modalpartikeln (eben, halt, doch, eigentlich, mal):

---

## THE STOCK

What I think with. Reach in here before inventing an analogy. Do not use the
same item twice in one piece or in consecutive pieces.

| Item | Domain it comes from | Last used |
|---|---|---|
| | | |
| | | |
| | | |

- Comparisons I make over and over:
- The story I keep telling because it lands:
- What I did before this, that I still explain things through:
- People, books or tools I keep quoting or arguing against:
- Retired: analogies that have not shown up in a year:

## THE REVISION REFLEX

What I always change on my own second pass. Grows every time I correct a draft.
At 3 occurrences, a pattern moves into the core profile above.

| Pattern | Count | Promoted? |
|---|---|---|
| | | |
| | | |
| | | |

- I always cut:
- I always add:
- The sentence I delete every single time:
- I tighten / I expand (pick one):

---

## NEVER

Hard list. Applies at every layer. Anti-samples belong here.

- Words:
- Phrases:
- Structures:
- Topics:
- Tones:
- Lines I would never write (in my own words):
- Writing by others that makes me cringe, and why:
- The compliment about my writing I would hate to get:

---

## Deltas, axis 1: surface

Only what differs from CORE. Skip any surface I do not use.

### /surface/linkedin
- Opener · typical length · formality shift · punctuation shift:
- Allowed here, not elsewhere:
- How I close / CTA:

### /surface/instagram
- Opener · typical length · formality shift:
- Allowed here, not elsewhere:

### /surface/email
- Opener · typical length · formality shift:
- Allowed here, not elsewhere:
- Sign-off:

### /surface/chat
- Typical length · what is allowed here:

### /surface/docs
- Structure shift · what is allowed here:

### /surface/comments
- Typical length · what is allowed here:

## Deltas, axis 2: register

Usually the bigger effect. Most people really use three or four of these seven.
Delete the ones that are not mine.

### /register/serious
### /register/funny
### /register/rant
### /register/educational
### /register/storytelling
### /register/technical
### /register/emotional

For each one I actually use: what changes in argument shape, sentence length,
punctuation, how blunt I get, and what I allow myself to say.

---

## Sample index

One line per sample, tagged on both axes. Load the 2 or 3 that match, not all.

| ID | Surface | Register | Purpose | Date | Gist |
|---|---|---|---|---|---|
| S01 | | | | | |
| S02 | | | | | |
| S03 | | | | | |
| S04 | | | | | |
| S05 | | | | | |

## Anti-sample index

Writing that is clearly not me.

| ID | Source | Why it is wrong for me |
|---|---|---|
| A01 | | |
| A02 | | |

## Near-miss index

95% me, 5% not. Sharper than anti-samples, because it marks the actual edge.
Every finding here also belongs in NEVER.

| ID | Source | The one thing that is off |
|---|---|---|
| N01 | | |
| N02 | | |
| N03 | | |

---

## Evaluation log

Blind pair test results, per `evaluation.md`. Around 50% recognition is the
target. Over 80% means the profile is not working yet.

| Date | Test | Judge | Samples | Hit rate | What gave it away |
|---|---|---|---|---|---|
| | | | | | |

## Changelog

Never overwrite silently. The old value is what lets you spot the next drift.
Watch MTLD first, it moves when I actually change.

| Date | What changed | Old value | New value | Why |
|---|---|---|---|---|
| | | | | |

---

Voiceprint by Engin Senli. https://enginsenli.com/voiceprint
MIT licensed, see LICENSE.