About

Writer's block costs hours. The average fiction author loses 15 - 30 productive minutes per session searching for a starting point. This generator constructs writing prompts algorithmically by combining G (genre), S (setting), C (character archetype), K (conflict type), D (narrative device), and M (mood) from a curated dataset of over 500 component entries. Each prompt is assembled using weighted selection that respects genre-setting affinity scores, producing coherent combinations rather than random noise. The complexity parameter L controls how many constraints layer onto the base prompt, from a simple single-sentence seed at L = 1 to a multi-layered scenario with narrative restrictions at L = 5.

This tool approximates creative direction, not creative output. It cannot replace an author's voice. The prompts assume familiarity with narrative structure. Edge case: at maximum complexity with narrow genre filters, combinations may feel overwrought. Reduce L or broaden your genre selection. Pro tip: use the seed-sharing feature to send identical prompts to writing group members for comparative exercises.

Formulas

Each prompt is assembled from n independent component pools. The total combinatorial space T for unique prompts at complexity level L is:

T = |G| × |S| × |C| × |K| × L∏i=1 |D_i|

where G = genre set, S = setting pool, C = character archetype pool, K = conflict type pool, and D_i = the i-th optional constraint layer (narrative device, mood modifier, structural constraint). At L = 1, only the base four components are used. Each additional level adds one constraint dimension.

Anti-repetition uses a recency buffer of size b = 5 per pool. If a candidate x appears in the buffer, its selection weight w is reduced: w_eff = w × 0.1. The seed-based pseudo-random number generator uses a Linear Congruential Generator: x_n+1 = (a ⋅ x_n + c) mod m, where a = 1664525, c = 1013904223, m = 2³².

Reference Data

Genre	Typical Conflict Types	Common Settings	Recommended Complexity L	Word Count Target
Literary Fiction	Internal, Relational	Domestic, Urban	3 - 5	3000 - 8000
Science Fiction	Technological, Societal	Space, Future City, Lab	3 - 5	4000 - 10000
Fantasy	Quest, Moral, Power	Kingdom, Wilderness, Ruins	2 - 4	5000 - 15000
Horror	Survival, Psychological	Isolated House, Forest, Hospital	2 - 4	2000 - 6000
Mystery / Thriller	Deception, Investigation	Small Town, Mansion, Office	3 - 5	4000 - 12000
Romance	Relational, Social	Café, Travel, Workplace	1 - 3	3000 - 8000
Historical Fiction	Societal, Moral, Survival	War Zone, Court, Colony	3 - 5	5000 - 15000
Dystopian	Rebellion, Identity, Survival	Megacity, Wasteland, Bunker	3 - 5	4000 - 10000
Comedy / Satire	Social, Absurd, Bureaucratic	Office, Suburbia, Online	1 - 3	1500 - 5000
Magical Realism	Internal, Existential	Village, Market, Family Home	2 - 4	3000 - 7000
Western	Justice, Survival, Territorial	Desert Town, Ranch, Train	2 - 4	3000 - 8000
Gothic	Psychological, Supernatural	Castle, Moor, Cathedral	3 - 5	4000 - 10000
Flash Fiction	Any (single focus)	Any (minimal)	1 - 2	100 - 1000
Poetry	Emotional, Sensory	Abstract, Nature, Memory	1 - 3	20 - 200 lines
Noir	Betrayal, Corruption	Rainy City, Bar, Alley	3 - 5	3000 - 8000
Slice of Life	Internal, Relational	Apartment, School, Park	1 - 2	1500 - 5000
Experimental	Structural, Meta	Abstract, Fragmented	4 - 5	Variable
Children's	Friendship, Discovery	School, Forest, Imaginary	1 - 2	500 - 3000

Frequently Asked Questions

Each genre has an affinity map assigning weight multipliers to settings. For example, Science Fiction assigns a weight of 3.0 to "Space Station" but only 0.5 to "Medieval Village". When the generator selects a setting, it multiplies each candidate's base weight by the active genre's affinity score, making coherent pairings far more probable. At complexity level L = 1, only high-affinity combinations appear. At L ≥ 4, lower-affinity settings can surface intentionally to create cross-genre challenge prompts.

Yes. The seed value initializes a deterministic Linear Congruential Generator with constants a = 1664525, c = 1013904223, m = 2³². Given identical seed values and identical filter settings (genre, complexity level), the output sequence is identical. Changing any filter parameter alters the pool sizes and thus changes the modular selection indices, producing different results even with the same seed.

The anti-repetition system reduces weights to 10% but never to zero. If a pool has fewer entries than the buffer size b = 5, some entries will always be in the buffer, but they remain selectable at reduced probability. The system guarantees termination. For pools with fewer than 6 entries, consider the recency buffer as a soft preference rather than a hard exclusion.

Both. At L = 1, the prompt contains only genre + setting + character + conflict - four tightly coupled elements. Each additional level adds one constraint layer: L = 2 adds a narrative device (e.g., unreliable narrator), L = 3 adds a mood modifier, L = 4 adds a structural constraint (e.g., 'must begin at the ending'), and L = 5 adds a thematic requirement. Higher levels produce longer prompts and demand more sophisticated narrative planning from the writer.

With approximately 18 genres, 45 settings, 30 character archetypes, 20 conflict types, 15 narrative devices, 12 moods, 10 structural constraints, and 14 thematic requirements, the space at L = 5 exceeds 18 × 45 × 30 × 20 × 15 × 12 × 10 × 14 = over 12 billion unique combinations. Even at L = 1, you get 18 × 45 × 30 × 20 = 486,000 base prompts. Repetition within a single writing career is statistically improbable.

The dataset skews toward Western literary conventions because the genre taxonomy derives from English-language publishing categories. Settings include global locations, but conflict archetypes follow predominantly Euro-American narrative theory (Freytag's pyramid, Campbell's monomyth). Writers working in non-Western traditions should treat the output as a structural suggestion and adapt cultural context independently.