Crypt Table's generators combine entries from fixed, project-specific game tables in your browser. They do not scrape another website, call a live AI model, or publish a new web page for every result. A generated NPC or tavern exists only on your screen unless you copy it into your own notes. This page documents the choices behind those tools so you can judge a result instead of treating a random answer as authority.
Some explanatory copy and source-list revisions were prepared with generative-AI assistance and checked against the current code before publication. That assistance is separate from using a generator: clicking a tool makes a local random selection from the fixed arrays already shipped with the page.
Random means a prompt, not a finished ruling
Each click chooses entries from a fixed table with JavaScript's Math.random(). That is suitable for casual tabletop play, but it is not a cryptographically secure draw and it is not meant for gambling, prizes or any decision with real-world consequences. Because entries are combined independently, the code cannot guarantee that every possible cross-table combination will read naturally. If two fragments clash with the scene, keep the useful part, edit it or roll again.
Most choices are uniformly random: every entry in one list has the same chance. The exceptions are stated below. The tools do not secretly rank results, infer anything about the player, or send the question or generated text to a server.
Names, characters and NPCs
The name generator draws from separate given-name and byname lists for six broad fantasy traditions. A personal name has a 70% chance to receive a byname or epithet. Town and tavern names join two fixed word banks, and a batch rejects duplicates before displaying up to twelve results. The labels describe fictional sound palettes, not real-world languages or historical naming advice.
The NPC generator selects a name, ancestry, occupation, visible detail, personality, mannerism, want and secret independently. Independence creates useful surprises, but it can also create tension: a quiet priest may receive a boisterous mannerism, or a wealthy-looking stranger may want a modest sum. Resolve that tension as characterization or reroll one whole profile. The tool supplies no game statistics and makes no claim that an NPC is balanced for a particular ruleset.
The character generator uses the common “roll four six-sided dice and drop the lowest” procedure for each of six abilities. Names, archetypes, backgrounds, traits, flaws and motives come from separate project-specific lists. The displayed ability modifier is floor((score − 10) ÷ 2). These conventions fit many d20-style games, but your rulebook decides whether to keep, rearrange or ignore them.
Places, treasure and encounters
The tavern generator combines a sign, keeper, drink, food, room mood, two distinct patrons and a rumor. Those details are deliberately modular. A practical way to use the result is to connect one patron to the rumor and decide what the keeper loses if the party investigates. That human connection turns a list of ingredients into a scene.
The loot generator has four fictional wealth bands. Coin ranges, valuable objects and magic-item chances rise by band. The gp, sp, cp and pp labels are familiar fantasy shorthand; their buying power is not tied to a published economy. “Hoard” does not mean appropriate for a certain character level. Compare the result with prices and advancement in the game you actually run.
The encounter builder calculates round(party size × (level + 1) × difficulty × 1.1), then spends that internal budget on creatures from one theme. It stops at fourteen creatures. Creature weights are fixed, project-specific relative values, not XP, challenge ratings or a safety guarantee. The setting and twist are rolled separately and are not included in the budget, so visibility, hazards, surprise and extra enemy actions still need review at your table.
Oracle probabilities
The solo oracle rolls from 1 to 100 and answers yes when the roll is at or below the likelihood you selected. At 50/50, exactly fifty results answer yes. Strong “and” results occupy roughly the outer fifth of the chosen yes or no range; ordinary answers sometimes gain a “but” complication. Doubles from 11 through 99 also trigger a random event. The on-screen roll and threshold stay visible so you can check the decision.
Privacy, saving and review
Generation happens locally. Stateless generators forget their output when the page closes. Trackers, the character sheet, journal and Gallowmere game use browser storage and are described separately on their pages. Browser storage is convenient, not a backup: clearing site data or changing devices can remove it. See the privacy policy for analytics and storage details.
If a fragment is unclear, duplicated, offensive, or inconsistent with the rule described here, send the exact result through the contact page. The rules on this page match the live tool code as of September 5, 2026. Corrections are applied to the source list so later rolls benefit from them.