124 Scenarios, 268 Voices: Inside Ghost RP
Most AI roleplay platforms optimise for speed. Ghostproof RP optimises for the sentence. The same editorial engine that powers our Books product runs on every RP response, in real time, with no manual editing possible and none needed.
The AI roleplay market has a prose problem. Platforms generate fast, conversational responses that read like chatbot output dressed in genre clothing. The fantasy tavern scene sounds like the cyberpunk interrogation scene sounds like the romance encounter. Strip the setting details and the underlying voice is identical. Competent. Functional. Forgettable.
Ghost RP was built to fix that specific failure.
What makes it different
Individual narrator voices. Each of the 124 scenarios has its own Voice DNA profile. The noir detective narrates differently from the supernatural anime guide. Sentence length distribution, clause complexity, register, metaphor domain, interiority ratio. These are not genre presets. They are prose fingerprints extracted from real writing and locked into the generation.
A western scenario narrator writes in short declarative sentences with dry observations and physical detail. A gothic horror narrator favours long, winding clauses that delay the reveal. A cozy mystery narrator maintains warmth while withholding information. 268 voices across 27 genres, each distinct enough that switching scenarios changes how the prose sounds at the sentence level.
The editorial engine runs on every response. The same 265+ constraint rules that power Ghostproof Books fire during RP generation. No em dash chains. No perception filters. No body-emotion sync. No narrator editorialising. No scene-ending summaries that restate what just happened. These patterns get caught and prevented before the response reaches the player.
Life Injection on every exchange. NPCs hold contradictory feelings. Their bodies react before their minds catch up. They have opinions about irrelevant things. They start thoughts and abandon them. The eight categories of involuntary human cognition fire on every generated response, making characters feel like people with inner lives rather than response generators following a script.
The genre range
27 genres. Not 27 variations on fantasy with different skin. The scenario library spans thriller, horror, supernatural, cyberpunk, western, cozy mystery, literary fiction, anime, noir, romance, historical, post-apocalyptic, political intrigue, medical drama, and more. Each genre built from the ground up with its own pacing expectations, tonal register, and narrative conventions.
Six of 124. The full library is browsable by genre, tone, and content rating on the RP page.
What actual output sounds like
That is not cherry-picked. The quality band across the 124 scenarios is narrow by design. We tested six scenarios during development and scored them on prose quality, voice consistency, and AI fingerprint count. The floor was 7.5 out of 10. The ceiling was 9. Zero ICK words across roughly 2,000 words of test output. The system is reliable, not lucky.
How authors use it
Some authors use Ghost RP to test story premises before committing to a full novel. Play through a concept as interactive fiction. See if the characters hold tension over ten exchanges. See if the world sustains interest. See if the emotional arc has legs. If the premise works in RP, it works in Books. If it collapses after three exchanges, you saved yourself a month.
The RP engine and the Books engine share the same editorial architecture. Voice DNA, constraint rules, Life Injection. A story that begins as an RP adventure can become a novel outline. The quality standard does not change between products.
The demo
Three scenarios. Ten exchanges each. No account. No email. No paywall. Enough to hear how a narrator voice sounds when it has been built from the ground up rather than prompted on the fly.
The demo runs the full editorial engine. Same 265 rules. Same Life Injection. Same Voice DNA. The only limit is the exchange count.
What comes after RP
Ghost Companion takes the same architecture in a different direction. Where RP gives you a narrator and a world, Companion gives you a single character with persistent memory, evolving trust, and a diary it writes about you when you are not there. The editorial engine prevents sycophancy. The companion has thoughts it has not volunteered and opinions it is still forming. Coming later in 2026.
Three products. One editorial engine. Books for publication. RP for play. Companion for connection. The prose quality is the same across all three because the architecture demands it.
Play the demo
Three scenarios. Ten exchanges. No signup. Enough to hear the difference between AI roleplay that prioritises speed and AI roleplay that prioritises the sentence.