How to Build a Party Dynamic with a Character Headcanon Generator

The four members are what the generator gives you; the three heavy lines are the only part you have to write yourself. Grey lines are the pairs you deliberately leave alone.

Most party advice is single-character advice repeated four times. This page is about the other half: the moments where the party moves as a unit, and who moves first when it does. Everything below is built on the character headcanon lines this site actually produces, and every count quoted here was measured from the bank inside the page source rather than estimated.

Why the generator can only ever give you members, not a party

Start with the constraint, because it decides the whole method. The bank on the generator page is 5 categories × 6 tones × 6 lines per cell = 180 hand-written lines. Every one of those 180 lines contains the placeholder {name} exactly once — measured: 180 of 180 lines hold one placeholder, and 0 of 180 hold two.

That single measurement has a large consequence. A line with one slot can describe one character. It cannot describe a relationship between two characters, because there is no second slot to put the other name in. No amount of rolling will produce "Marlow and Priya have a standing argument about who pays", because the sentence pattern has nowhere to put Priya.

The group is barely present in the wording either. Of the 180 lines, only 4 acknowledge that the character belongs to any group at all — a group chat left unread, being the unofficial keeper of the group's birthdays, being the reason the group chat needed a second group chat, and still saying we about a group that has not met in years. Three of those four sit in the Relationships category, which means 33 of the 36 relationship lines point at one specific person rather than at a crowd.

The interface agrees with the bank. There is one name field, and the whole page keeps a single state.name while you roll. One character in, one character out.

So treat this as a division of labour rather than a shortcoming. The generator's job is the individual: the quirk, the private thing, the way they behave under pressure. Your job is the group. Once you accept that, you stop re-rolling in the hope that a party will fall out of the tool, and you start spending your effort where it actually changes the result.

The arithmetic: how many edges you just signed up for

A party is not a list of members, it is the edges between them. The count grows faster than the roster does:

• 3 members — 3 edges
• 4 members — 6 edges
• 5 members — 10 edges
• 6 members — 15 edges
• 7 members — 21 edges

This is why parties feel thin long before they feel crowded. Four members already means six relationships to keep straight, and fifteen is past the point where anyone can track them live. The instinct is to describe everyone more thoroughly. The fix is to write fewer edges on purpose.

Write three. Not three per member — three total, for the whole party, whatever its size. Three is small enough to hold in your head during a scene and large enough that the group has a shape when it acts. A four-member party with three defined edges already reads as more organised than a six-member party with none.

The three edges worth writing

Pick one of each. The categories matter more than which specific members you pick.

• Allegiance — two members who end up on the same side, and the thing they are on the same side about. This is the edge that lets the party make a decision without you narrating a discussion.
• Friction — two members who disagree about the same thing every time, and neither of whom will move. One recurring disagreement is enough. Two is a party that cannot function; zero is a party with no texture.
• Silence — two members who never address each other directly, plus the wrong reason one of them thinks that is. The silence is what the players notice first, because the absence of an interaction is louder at a table than any interaction.

Each edge needs exactly one observable behaviour attached to it. Not a feeling, not a summary of the relationship — a thing that happens that the players can see or hear. "Ines and Otis agree on the plan" is a feeling. "Ines starts packing, and Otis is already holding the list" is an observable behaviour, and it took the same number of words.

If you cannot state an edge as something observable in one sentence, you have written a description of a relationship instead of the relationship itself, and it will not survive contact with a scene.

Rolling a four-member party in five steps

• 1. Roll member one with the panel wide open. Leave all six tones on. Write down the three lines, and write down the Archetype label printed above them — that label is one of 24 and it is drawn before the lines, from the same seed.
• 2. Change the name and roll again. Do not reuse the previous name to "keep some continuity" — you will not get continuity, you will get a different batch. The draw is a hash of the name plus the tones, categories, count and roll number, so a new name reshuffles everything. In our own four-member test, all six member-to-member pairs were compared line by line and 0 of 18 lines repeated. Members do not contaminate each other, which is exactly what you want.
• 3. Write the four Archetype labels in a row before reading any lines. You are checking for a collision, and a collision is much easier to see when the four labels sit next to each other than when they are buried in the output.
• 4. Run the collision check described in the next section. Decide what a repeated label means before you start writing.
• 5. Read each member's lines as evidence about the party, not about the member. This is the step that does the actual work:
  • Their Relationships line tells you who they orbit. Orbit is an edge.
  • Their Secrets line tells you what the others do not know. That gap is an edge.
  • Their Quirks line is what the rest have learned to live with. Tolerance is an edge.
  • Their Personality line is what shows when the party is under pressure. Pressure is where edges become visible.
  • Their Backstory line is usually the one you cut. A party of four does not need four histories; it needs four people whose present behaviour rubs.

Four members × three lines gives you twelve candidate observations. Three of them become edges. The other nine stay as background, and that is the correct ratio — a party where all twelve observations are load-bearing is a party that cannot be improvised with.

Two members will land the same Archetype about a quarter of the time

The 24 Archetype labels are drawn from the same random stream as the lines, before the lines are picked. They are not checked against each other, and there is no de-duplication hiding in the code. Measured over 20,000 random four-member parties:

• four different labels — 76.26%
• three different labels (one pair collides) — 22.60%
• two different labels (two collisions) — 1.15%

At least one collision therefore happens in 23.74% of four-member parties. For reference, four independent uniform draws from a set of 24 would collide 23.13% of the time — so the label behaves almost exactly like plain chance, with no hidden de-duplication. If you roll four members, expect a shared label roughly one time in four, and expect to have to make a decision about it.

Our own worked party below is one of those cases. Four names, six rolls in, and Ines and Otis both came out The Overprepared.

The wrong response is to re-roll until the labels are clean. A shared label between two members is free material: it is either an allegiance (these two agree about method) or a rivalry (these two agree about method and cannot both be in charge of it). Both are more interesting than a tidy set of four, and both take one sentence to write.

Re-roll only when the shared label actively contradicts a line you have already committed to. Otherwise keep it and pick which of the two readings you want.

Locking one tone for the whole party makes members repeat each other

There is a second, less obvious consequence of how the bank is built, and it bites when you want the party to share a mood.

With all six tones open, each category has 36 lines to draw from and the whole pool is 180. Lock the panel to a single tone and each category cell drops to 6 lines, so the entire pool falls to 30 — six times smaller. Members are then drawing from the same small slice, and they start to look like each other:

• All six tones open — 26.78% of four-member parties contain at least one pair of members who drew an identical line. Average repeated pairs per party: 0.299 out of 6.
• One tone locked — 87.77% of four-member parties contain at least one such pair. Average repeated pairs: 1.658 out of 6.

Locking a single tone raises the chance that two members overlap by roughly 3.3 times. The scene you wanted was "the whole party is grieving"; what you get is two members who are grieving in the exact same sentence, which reads as an author who ran out of ideas rather than as a group sharing a mood.

So put the shared mood in the scene, not in the panel. Keep all six tones on while rolling, and make the party's common feeling something you state yourself — the situation they are in, not the register the sentences were written in. Even with all tones open, note that roughly one party in four will have two members overlapping on a line, which is worth one glance before you commit.

A worked party, annotated

Here is a real four-member party, reproducible on the generator page: four names, all six tones on, all five categories on, three lines each, first roll. The lines below are copied exactly as the generator produced them, with only the placeholder replaced by the character name.

Marlow — Archetype: The Patient Explainer

• Secrets / Wholesome — "Marlow keeps every thank-you note they have ever received locked in a drawer, sorted by year."
• Quirks / Angsty — "Marlow rehearses conversations that have not happened and probably will not happen."
• Backstory / Angsty — "Marlow learned early that bad news travels best when delivered calmly, so they deliver everything calmly."

Marlow is filed by the party as the calm one, because that is what the Backstory line says and it is the most visible of the three. The Quirks line says something else: the calm is preceded by rehearsal. Nobody in the party knows that, and the Secrets line explains why nobody would — a drawer of thank-you notes is not a thing you show people.

Ines — Archetype: The Overprepared

• Backstory / Romantic — "Ines moved cities once for a person, and would tell you it was for the job."
• Personality / Chaotic — "Ines agrees to everything for a week, then disappears to recover from the calendar they built."
• Relationships / Funny — "Ines argues about restaurant choice for twenty minutes, then orders whatever their friend orders."

Ines is the party's planner, and the two lines that matter for a group are the second and third. The second is a pattern the rest of the party has already noticed. The third tells you Ines argues for the sake of arguing and then defers — which is a very different thing from wanting to be in charge, and is exactly the kind of distinction a player will pick up on if you let them see it twice.

Otis — Archetype: The Overprepared

• Quirks / Angsty — "Otis counts the exits in every room without meaning to."
• Backstory / Funny — "Otis spent two years believing a family legend, and still half-believes it when tired."
• Relationships / Romantic — "Otis notices when the person they like has changed their hair before anyone else in the room."

Priya — Archetype: The Slow Realist

• Relationships / Chaotic — "Priya makes plans for eight people and is genuinely surprised when eight people arrive."
• Personality / Chaotic — "Priya has three plans running at once and no idea which one is currently winning."
• Quirks / Angsty — "Priya eats too fast, from years of meals that came with a time limit attached."

Twelve lines, four people, and not one of them is a statement about the group. Now the assembly, using the three slots:

• Allegiance — Ines + Otis. Both rolled The Overprepared. Read it as agreement about method: Ines builds the calendar, Otis checks the room. They are always on the same side of "we should have planned this", and the party has stopped arguing with the pair of them about it.
• Friction — Ines + Priya. Ines agrees to everything for a week, then disappears; Priya makes plans for eight people and is genuinely surprised when eight people arrive. These two lines conflict directly. Ines is quietly managing a capacity Priya keeps exceeding. The observable behaviour: when Priya starts inviting people, Ines stops answering for about three days, and Priya has never connected the two.
• Silence — Marlow + Priya. They never speak to each other directly. Priya has decided this is judgement, because Marlow is calm and Priya is not. It is not judgement — Marlow rehearses conversations that have not happened, and this is one of them. Marlow has the line ready; they have simply never delivered it. That gap is the edge, and it costs one sentence to keep alive.

Three edges from twelve observations, using nothing that was not already on the page. Note that all three came out of lines that were individually about a single person — the group reading is entirely in which pairs you choose to connect.

A fill-in sheet for any party

Once the four members are rolled, the writing takes one line per edge. Copy this and fill the blanks:

• Allegiance: ______ and ______ are always on the same side about ______.
• Friction: ______ and ______ disagree about ______ every time, and neither has moved.
• Silence: ______ never speaks to ______ directly, and ______ thinks it is because ______.

Two rules keep the sheet honest. First, every blank must be fillable from something the players could observe — no blank may be filled with a feeling. Second, if you cannot fill a blank from the lines you rolled, do not invent the missing fact; move the edge to a different pair. The twelve observations usually contain a better edge than the one you were reaching for.

What to cut

• Three of the four Backstory lines. Keep at most one, and keep the one whose behaviour is already visible in the other lines.
• Any edge you cannot state in one sentence. If it needs a paragraph, it is a plot, and plots belong in the campaign rather than in the party's standing shape.
• The second disagreement. One friction edge holds a party together interestingly. Two of them make the party unplayable, and the players will start avoiding both members rather than leaning into them.
• Perfect balance. A party where every member has an equal share of screen time has no shape. The three edges should be unevenly distributed — one member appearing in none of them is a legitimate and useful outcome.

Putting the three edges to work

The test of a party sheet is whether it can survive a scene you did not plan. Set up a decision the group has to make together and read your three edges against it: does the allegiance produce a block of two, does the friction produce a holding pattern, and does the silence produce one member who says nothing when everyone expects them to?

If all three fire, the party is doing the work and you can let the scene run. If none of them fire, the edges are descriptions rather than behaviours, and the fix is not more characters — it is rewriting the three lines so each one names something that happens.

How this relates to the rest of the site

The generator is one HTML file with a fixed bank of 180 hand-written lines and no model of any kind. If you want the category and tone layout before you start rolling, the 30-cell matrix sets out what each of the five categories is for. If you want to see how one member changes across all six tones, the six-tone comparison does that with a fixed name. If your party members are non-player characters rather than protagonists, the tabletop NPC guide covers how many lines each one actually earns. And if you want to know why rolling the same settings eventually repeats, the repeat analysis covers the pool arithmetic.

Quick answers

Can the generator make a party dynamic on its own? No. Every one of the 180 lines contains one character placeholder, and 0 of 180 contain two, so no line can refer to a pair. The generator produces members; the group is assembled by hand.

How many edges should a four-member party have written? Three. Four members have six edges, and three is the number you can keep in mind during a scene while still giving the group a shape.

Two of my members rolled the same Archetype. Should I re-roll? Usually not. It happens in 23.74% of random four-member parties, which is close to what plain chance would produce from 24 labels. Treat the shared label as either an allegiance or a rivalry, and only re-roll if it contradicts a line you have already committed to.

Should I set one tone for the whole party? No, and this is the one setting that actively hurts. Locking a single tone shrinks the pool from 180 lines to 30, and the share of four-member parties containing two members who drew an identical line rises from 26.78% to 87.77%. Keep all six tones on and put the shared mood in the scene.

Do I need all five categories for a party? Roll with all five, then cut. Backstory is the category the group needs least — at most one member should carry a past, and only if that past already shows in their behaviour.

Is any of this uploaded or stored? No. The lines come from a bank shipped inside the page, the name is only used to fill the blank, and there is no account, no storage and no network request. The privacy page and the offline test agree on that.