Skip to content

Visual Direction Brief — Scenario Imagery

For: Design partner exploring AI-generated scene-setting imagery for Sync From: Jonathan / Sync team Status: Open exploration — no constraints on style direction yet

This brief is designed to be pasted into an AI assistant (ChatGPT, Claude, etc.) as context, then iterated on with the designer driving. The goal is for you to come back with a visual direction + a reusable prompt template + sample renders, not finished imagery for production.

What Sync is

Sync is a decision-making game. Players are dropped into a scenario — historical (Cuban missile crisis, Roman senate), fictional (pirate mutiny, superhero ethics dilemma), modern (boardroom layoff, startup co-founder dispute), speculative (post-apocalyptic bunker, space station crisis) — and choose how to act. Their choices are scored against the group's emergent pattern. That score, "Sync," is the core metric.

Two formats matter for imagery:

  • Quick games — one self-contained scenario, ~5–10 minutes. One image needed.
  • Campaigns — multi-chapter arcs in a single setting (e.g., 5-chapter pirate story, 5-chapter Cold War story). A series of images that need to feel like episodes of the same show.

Settings span anything: pirates, Romans, superheroes, WW2 codebreakers, medieval abbeys, modern Silicon Valley, sci-fi colonies, fantasy courts. There is no house genre.

What the image needs to do

It opens the scenario as a banner above the decision text. Within ~1 second the player should know:

  • Era (when)
  • Place / role (where, who)
  • Mood (tense, contemplative, chaotic, ceremonial)

Then they read the scenario text and decide. The image is supportive, not the main event — it sets the stage and gets out of the way.

What we need from you

  1. A visual style direction that survives radically different settings. The same product is asking you to render a Roman senate and a superhero rooftop and a startup boardroom — and they need to feel like they belong in the same app without looking identical.

  2. A reusable prompt template with placeholders that engineering can fill in programmatically per scenario. Roughly: [SCENE], [ERA], [MOOD], <your style anchors>. The style anchors are the load-bearing part — the variable inputs come from the scenario itself.

  3. Sample renders — 8–10 across the genre range, proving cohesion. We've put a starter list at the bottom of this brief.

  4. Aspect ratio recommendation. Current UI assumes letterbox banner (16:9 or 21:9). Open to other framings if you have a strong opinion.

  5. A fallback aesthetic for when generation fails or produces something off — what should the placeholder look like? Today it's a flat gradient.

Constraints to design around

  • Cheap model at scale. Engineering plans to use Replicate's FLUX schnell (~$0.003/image) for most generation. It's fast and cheap but stylistically inconsistent unless the prompt anchors it hard. Style cues that lean into "cinematic still, muted palette, film grain, soft focus, painterly" tend to survive cheap models better than ones that demand fine photorealistic detail. The style needs to look intentional even when the model is rough. If you make a strong case for a more expensive model, we'll listen — but the default assumption is cheap-and-anchored.

  • No identifiable real people. Scenarios reference historical figures (Kennedy, Caesar, etc.) but images should suggest era and role through costume, setting, and silhouette — not portrait likenesses.

  • No atrocity imagery. Per our content policy, scenarios don't use atrocities as decision substrate. Images follow the same rule — no graphic violence, war crimes, or distress imagery. Tense and consequential is fine; gratuitous is not.

  • Mood over accuracy. A pirate scene doesn't need historically accurate rigging. It needs to feel like a pirate scene in two seconds.

  • Text-friendly composition. The banner sits above scenario text. High-contrast subjects dead-center will fight the text for attention. Compositions with negative space, atmospheric depth, or off-center framing work better.

  • Cohesion within a campaign. A 5-chapter pirate campaign should feel like five stills from the same film. If your template depends on randomness, build in a way to seed-lock or style-lock per campaign.

Deliverables

  • Prompt template (or 2–3 variants if different scene types — interior/exterior, character-focused/landscape — warrant different treatments).
  • 8–10 sample renders across genres, showing the style holds.
  • Model recommendation — confirm FLUX schnell works, or argue for an alternative with a cost/quality trade-off.
  • Fallback aesthetic — one paragraph + a mock if relevant.
  • Notes on what doesn't work — if you tried directions that broke down, tell us. Saves us re-discovering them.

Test scenarios (mix-and-match — pick 8ish)

If your template can make these feel like they belong in the same product without all looking identical, we're in business:

  • A Roman senator deciding whether to back Caesar's land reform
  • A pirate quartermaster facing a captain's mutiny mid-voyage
  • A 1962 Kennedy advisor weighing a naval blockade in the Situation Room
  • A superhero choosing whether to break protocol to save a single civilian
  • A modern startup founder deciding which co-founder to fire
  • A WW2 codebreaker debating whether to act on intelligence that would expose the source
  • A medieval abbess negotiating with a visiting bishop
  • A space station commander rationing oxygen during a slow leak
  • A 19th-century newspaper editor deciding whether to publish a damaging story
  • A post-apocalyptic settlement leader weighing whether to take in refugees

What success looks like

When a player opens a scenario, the image lands before they read the text. They know the era, the stakes, and the tone. Then they read. Then they decide. The image fades from their attention because the decision is the thing.

We're not trying to make a game that's pretty. We're trying to make a game where the visuals do their job in a second and step aside.

Out of scope (don't worry about these)

  • UI/page design beyond the banner placement
  • Animation or motion
  • Branding, logo, marketing imagery
  • Localization of imagery
  • Per-character portraits or avatars

If you want to flag opinions on any of these, great — but it's not what we're asking for in this round.