D&D Character Generator

The group chat has been planning for three weeks. The DM has worldbuilding docs with a table of contents. And you, dear adventurer, have spent those three weeks deciding what accent your character will speak in, and zero minutes choosing anything else. Session zero is at six tomorrow evening. There are five classes you are considering, three races, and a growing suspicion that you will show up with 'a fighter, probably, whatever' and spend the first combat apologizing.

Or maybe the opposite problem: you are the DM of a one-shot, you need six NPCs statted by tonight, and the thought of hand-generating another bandit captain's stat line makes you want to invent a bandit captain just to rob someone else's prep time. Or you are a forever-player curious about the challenge-run scene, where you roll everything — race, class, the works — and are legally obligated to play the disaster the dice hand you.

This D&D character generator covers all three cases. One click assembles a complete fifth-edition character — race, class, background, ability scores by your chosen method, level-1 hit points, starting equipment, a name, and a one-line backstory hook. Then the part that makes it a tool instead of a slot machine: lock whatever you like, re-roll whatever you do not, and converge on the character you actually want in about ninety seconds. Free, instant, nothing to install.

A full generation produces the whole level-1 chassis: race with traits, class with features, background with its skill and tool proficiencies, six ability scores with modifiers already computed, hit points and Armor Class, starting equipment from the class list, a generated name, and a backstory hook — one specific, playable sentence like 'left the city watch after refusing a corrupt order' that is more useful than three paragraphs of orphanage.

• Race and subrace with traits, speed, and size
• Class and subclass direction, with level-1 features and hit die
• Background with proficiencies and the personality tables' greatest hits
• Six ability scores rolled by your chosen method, modifiers pre-computed
• Hit points, Armor Class, initiative, and saving-throw proficiencies
• Starting equipment and gold alternative
• A generated name and a one-line hook you can actually roleplay

Everything is plain 5e — the same rules vocabulary your table already speaks, drawn from the systems the Player's Handbook and the free basic rules document. No homebrew surprises, no third-party classes with three paragraphs of bespoke mechanics your DM has to adjudicate at 11 p.m. The names, for the record, come from the same naming engine as our Fantasy Name Generator, which is worth a visit when you want to browse names by culture and tone instead of taking what the character roller hands you.

1. 1

   Choose your stat method

   4d6 drop lowest for classic rolling, the standard array for fairness, point buy for control, or 3d6 in order for the old-school pain scenario. The method applies before anything else generates.

2. 2

   Set the randomization scope

   Full random for challenge runs, or constrain by category — classes you would enjoy, races that fit the campaign, any combination. The tool respects your constraints; it only fills the gaps.

3. 3

   Generate

   One click produces the complete sheet. Read the backstory hook first — it is the sentence that decides whether this character is interesting.

4. 4

   Lock what you like

   Every field has a lock. Keep the dwarf, the fighter, the 17 Strength, the name you cannot stop saying aloud. Locked fields survive every future re-roll.

5. 5

   Re-roll the rest

   Click again. Everything not pinned regenerates. Repeat until the sheet feels like a person rather than an assignment.

6. 6

   Print or export

   The sheet exports as a clean printable and as text you can paste into a virtual tabletop or a notes app. Session zero: solved before the kettle boils.

The entire culture war of 5e character creation fits in one question: how do you produce six ability scores? There are three sanctioned answers, plus the old-school fourth, and each embeds a philosophy. The basic rules lay them out plainly; the interesting part is the arithmetic underneath, because the numbers explain the arguments better than the arguments do.

Plain 3d6 averages 10.5 per stat — perfectly average farmers, which is what the 1970s game wanted, with six of those in order producing a total expectation of 63 and no reassigment allowed. The modern default, 4d6 drop lowest, averages about 12.24 per stat: rolling four dice and discarding the worst one skews the whole distribution upward, giving a six-stat expectation near 73.5. The standard array — 15, 14, 13, 12, 10, 8 — totals exactly 72, which is to say, deliberately a hair under the rolled average. Point buy caps you at a purchased 15 before racial bonuses but guarantees the floor. The designers, in other words, knew precisely what they were doing.

The variance is where the feelings live. A 4d6-drop-lowest set can produce a 6, and it can produce three scores of 16 or better — the probability of rolling an 18 on a single stat is 21 out of 1,296 (about 1.6 percent), a 17 is 54 out of 1,296 (4.2 percent), and a 16 is 94 out of 1,296 (7.3 percent), so roughly one stat in eight lands at 16 or above. One player at the table rolls like a legend, another rolls a librarian, and the array suddenly looks very civilized. Both experiences are features; only one of them is fair.

Let us roll one stat where you can watch every die. Four six-sided dice come up 6, 3, 4, and 6. We discard the 3, add the remaining 6 + 4 + 6, and record 16. That is the entire mechanic — here is a full six-stat block with the dropped die shown, then the assignment a player would actually make.

4d6 drop lowest, six stats:
  [6,3,4,6] drop 3 → 16
  [5,5,2,1] drop 1 → 15
  [4,4,4,2] drop 2 → 12
  [6,2,2,5] drop 2 → 16
  [3,1,3,5] drop 1 → 11
  [4,4,1,2] drop 1 → 10
  total: 80 — a strong roll, top-decile luck

Assignment for a hill dwarf fighter:
  STR 16 (+3)  DEX 10 (+0)  CON 16 (+3)
  INT 10 (+0)  WIS 15 (+2)  CHA 11 (+0)
  +2 STR / +2 CON from dwarf (pre-Tasha's custom lineage applied here)
  HP at level 1: 10 (d10 fighter) + 3 (CON) = 13

That total of 80 is genuinely lucky — the expectation was about 73.5, and 80 will outshine every array character at the table. Now the interesting decision: with those numbers, the obvious play is the fighter above, but a 15 Wisdom and 16s to spare would make a formidable cleric, and the generator's re-roll button exists precisely so you can explore that fork before committing. Lock the stats, re-roll the class, see what a warlock with these numbers looks like, then decide with your gut instead of your fear.

For the record, the array version of the same fighter: assign 15 STR, 14 CON, 13 DEX, 12 WIS, 10 CHA, 8 INT, apply the same racial bonuses, and land at HP 13 with a slightly worse to-hit. Different numbers, same character concept — the sheet flexes, the dwarf does not.

Every table argues this every few years, so here is the condensed version of the whole debate, with the recommendation column doing the adjudicating.

One nuance the table compresses: point buy's cost ladder is non-linear — 13 costs 5 points but 14 costs 7 and 15 costs 9 — so the efficient frontier is usually two or three stats at 14-15 and the rest at the 8-12 floor, exactly how the optimizer community builds. The generator's point-buy mode enforces the ladder and the 27-point budget automatically, including the classic error of forgetting that scores start at 8, not zero.

Here is the fact that settles a hundred forum arguments: price the standard array on the point-buy ladder and it sums to exactly 27. The array is not a lesser sibling of point buy or an approximation of it — it is a point-buy build, one specific allocation of the full 27-point budget, published as a shortcut for tables that would rather not do the shopping.

Once you see the array as one build among many, the interesting question becomes what else the same 27 points can buy — and the ladder's non-linearity makes the answer non-obvious. A flat build of 13, 13, 13, 13, 13, 10 costs the same 27 and totals 75, three points higher than the array's 72, because the ladder punishes the climb from 13 to 15. A spike build of 15, 15, 13, 10, 10, 8 also costs exactly 27 and totals only 71, trading raw points for a second +2 modifier. And a balanced 14, 14, 14, 13, 9, 8 — the three-fourteens build optimizers like for MAD classes — totals 72 with three +2 modifiers instead of the array's two.

That last observation is the quiet design masterstroke: across every sensible 27-point build, the sum of modifiers stays pinned around +4 to +5, because the price jumps from 13 to 14 and 14 to 15 land exactly on the modifier breakpoints. The game's economy charges you full freight for every +2 you concentrate. So the array-versus-point-buy debate was never about power — both spend the same budget at the same exchange rate — it is about who chooses the shape: the book (array), or you (point buy), or the dice (rolling, with variance as the house's cut).

Pure random generators produce characters; this one produces yours. The mechanism is the lock. Every field pins independently, so the workflow becomes: roll, examine, pin the two or three things that made you lean forward — the race, the backstory hook, the one perfect stat — and re-roll the rest. Three or four cycles of that and something odd happens: the character stops feeling generated and starts feeling discovered, the way a good hand-built character does, except this took four minutes instead of an evening.

DMs use the same loop differently: generate twelve NPCs, lock nothing, and cherry-pick the five with hooks worth keeping. One-shot authors pre-build the whole tavern that way. And the challenge-run community, of course, locks nothing ever, on principle, and plays the gnome wizard with 8 Constitution that the machine dealt them, with witnesses. There are no wrong workflows, only wrong witnesses.

A generated sheet with a blank name line produces analysis; a generated sheet with 'Bryn Hallowfen, dragonborn paladin who failed her order's final vigil and is seeking a worthier oath' produces a character. The backstory hooks are written as single playable sentences — a debt, a broken promise, a professional embarrassment — because a hook you can deploy in the first session is worth ten pages of genealogy you will never mention. Players who want deeper naming control should browse the Fantasy Name Generator by culture and tone, and writers building non-D&D casts have the general Character Name Generator for the same job in other genres.

The equipment block is likewise practical rather than exhaustive: the class kit from the rules, gold alternative, and the armor that makes your AC honest. It is the loadout your level-1 self can actually carry, not the shopping list you will argue about for an hour — that argument belongs to level 2, where tradition has always placed it.

It does not level past the early levels — it builds characters, it does not run a campaign's arithmetic — and it does not make optimization judgment calls for you beyond sane defaults. It is rules-faithful, not rules-lawyering: where the rules leave choices open (fighting style, cantrips, skill picks), the generator proposes and you dispose. It also cannot know your DM's house rules, which at some tables modify a quarter of the sheet, so treat the export as a strong first draft with the DM's red pen expected.

And it cannot want anything. The eternal truth of generators: the machine deals the cards, but caring about the hand is still your job. The lock button exists because we noticed that caring arrives somewhere around the third re-roll, and we wanted to make sure it had somewhere to stand.

• Procrastinating players, the night before session zero, in numbers that spike every time a new campaign forms
• DMs mass-producing NPCs and one-shot rosters faster than they can say 'the stat block is in the book'
• Challenge-runners who lock nothing and play what falls, posting the results as content
• New players using generation as a tour of the game's possibilities before they know what they like
• Writers mining random sheets for protagonists — a rolled character is a free writing prompt with dice
• Returning players from the 2e era, checking whether 4d6 drop lowest still feels the same (it does)

Character generation is pure local computation — dice, tables, and templates that fit comfortably in a browser tab — so that is exactly where it runs. No account, no uploads, no campaign data transiting anyone's server. Roll a hundred characters on a train with no signal and the tool will not even notice. The dice are Math-random-based with fair rolls, and the arithmetic is on the page where you can check it.

Pair the generator with the D&D Dice Roller for every roll the session itself demands, the Fantasy Name Generator when a name needs to be chosen rather than dealt, and the Character Name Generator for the NPCs your backstory hook inevitably drags into the plot. Character creation, dice, names — the whole stack lives in the browser and costs exactly nothing, leaving your budget for the important things: snacks, and more dice you do not need.