Most homebrew campaigns die of over-preparation. Someone writes forty pages of history for a continent the party never visits, runs out of energy before session two, and quietly stops scheduling. The method below is built to avoid that: prepare a situation, not a story, and build only as far as the party can walk in one evening.
A plot says what will happen. A situation says what is true, and lets the players decide what happens. Plots break the moment a party does something unplanned, which is immediately. Situations bend.
Write one sentence with instability built into it. "The mine that pays everyone's wages has flooded, and the only person who knows why has left town." That sentence already implies a location, an economy, a missing person and a deadline. Nothing in it dictates what the players do.
Test it: if the party does nothing for a month, does the world change? If the answer is no, it is not a situation yet.
This is the engine. Plot is what happens when incompatible goals collide, so you do not have to invent events — you set up pressure and let it out.
Three is the useful number. Two is a tug of war; four is bookkeeping. For each, write only: what they want, what they will do to get it, and what they are afraid of. Two lines each is enough.
Make at least one faction sympathetic and at least one internally divided. A faction the party might join is worth more than a faction they can only fight.
Six is roughly what a first session can hold. Each gets a name, one want, and one physical detail you can play at the table — a stutter, a limp, a habit of counting coins while talking.
The detail is not decoration. It is the thing that lets you become that person instantly when a player unexpectedly turns to them, and the thing players use to remember who is who.
Write the want as a verb: wants to buy back her father's debt, not is a merchant. Occupations sit still; wants create scenes.
A reliable shape for one evening: a hook that reaches the party where they are, two or three obstacles of different kinds — a fight, a conversation, a problem — a reversal that changes what they thought was true, and a payoff that answers one question while opening another.
Do not prepare session two. You do not yet know what your players will care about, and after one night you will know exactly.
Detail the village, the road out and the one place the hook points to. Everything else is a name on a horizon.
Blank space is not laziness; it is where the campaign will actually grow, because it grows toward whatever the players poke at. Filling it in advance means throwing most of it away.
A page. Genuinely. The situation, three factions at two lines each, six names, and the shape of one session. If your prep does not fit on a page, you are writing a novel, and a novel is a document your players will never read.
Everything above is the work. If you want the campaign without the page, that is what a generator plus an Autonomous Dungeon Master is for. MONOLITH builds the situation, the factions and the first episode from a description you give it, then runs the game — 5e compatible rules, real dice, one to six players. Established names and facts are recorded as they come up and fed back every turn, so the world stays consistent as it grows.
The first session is free and solo, and runs 24 rounds — about an evening.
Also:
what an AI homebrew campaign generator can and cannot do →
Also: building a 5e character →