Skip to content

Preamble Cinematic Redesign — Brief

Status: Brief — not an ADR. Architectural decisions (image-gen API, caching strategy, BYOK flow) will be filed as ADRs once scoped.

Date: 2026-05-15

Problem

The current chapter preamble renders as a single scrolling block of text. Playtesters (Cici, May 15) describe it as a wall of text that misses the chance to set the scene the way a video-game cinematic does before gameplay starts.

Goal

Replace the scrolling preamble with a clickthrough cinematic of 5–10 image + paragraph cards. Each card pairs one short paragraph with one image. Player taps/clicks through at their own pace, then enters the chapter.

Constraints

  • Image-gen cost is the binding constraint. Per-play image generation is not viable at scale.
  • Pre-generate for canned stories (one-time cost, amortized across all plays of that story). Store images alongside arc/chapter content.
  • Defer world mode (player-generated worlds) until BYOK or a tokenized model lands. Reference: Pax Historia precedent in project_monetization_roadmap.md.

Open questions

  • Image style consistency across cards within a story (and across chapters). Single style guide per arc, or per story?
  • Paragraph splits — LLM-authored from the existing preamble text, or human-curated by the story author?
  • Skip affordance — yes, but skips should be tracked as a behavioral signal (compare to existing "interesting" / "doesn't get me" reactions).
  • Mobile layout — swipe between cards vs. tap to advance. Default to tap (matches existing flow); swipe as enhancement.
  • Re-watch — if a player returns to a chapter mid-campaign, do they see the preamble again? Probably skippable-but-available.

Out of scope

  • Chapter scenes themselves (only the preamble before Chapter 1, or before each chapter — TBD).
  • World-mode preambles (deferred to BYOK).
  • Audio / music / motion beyond static image transitions (could be a v2).

Source signal

  • Cici's playtest feedback, May 15, 2026 ("make this part a bit more visual… click through five to ten images").
  • Jonathan's comparison to video-game cinematics: "this is what I would, this is it. This is that part."

Next step

Spin up an ADR when the first real fork appears — likely image-gen API selection (Replicate vs. fal vs. native Anthropic image-gen) or the canned-vs-BYOK split mechanism.