Till the Thaw
Branching dialogue for an AA narrative survival adventure, where you feel the weight of a choice without ever seeing the math (an NDA-safe rebuild of a real client project)
Role: Game UI/UX Designer · Timeline: 2024 (client project), 2025 (NDA-safe rebuild), 2026 (UE5 build) · Team: Client project (NDA) · Platform: PC / Console
Narrative UI, Dialogue System, Survival UX, AA Production
Overview
Till the Thaw is an AA narrative survival adventure set in a dying mining town in 1897. Resources are scarce, alliances are shaky, and a conversation can decide as much as a bullet. I designed the UI systems around one promise: the player should feel the tone and consequence of every choice in their gut, not read them off a stat sheet. This is an NDA-safe rebuild of a real client project. The thinking and the process are real. Every screen and name was rebuilt from scratch, and the numbers are reconstructed to keep the pattern of the real findings.
The Challenge
Branching dialogue comes with an ugly trade-off: honest systems look like spreadsheets, and pretty ones hide what matters. The brief was to make every reply communicate two invisible things, its emotional tone and the weight of its consequences, without exposing the system underneath. No “+15 trust” popups. No red “this will have consequences” banner doing the feeling for the player. And around the dialogue sat a full survival layer (vitals, navigation, crafting) that had to stay readable over painterly low-light environments without burying the world under panels.
The Solution
Every dialogue option carries a tone signature (Empathy, Defiance, Caution, Pragmatism), shown through color, icon and the voice of the line itself, plus a small row of pips that hints how far the choice will echo. You can tell a knife from an olive branch at a glance, and the math stays backstage. Under the screens there's a full information architecture (five domains, shared state, cross-system dependencies) where UI views read progression data but never own it. The HUD follows the same restraint, and I didn't get there by taste alone. The first iteration went through a moderated playtest, the findings became a written audit (what works and why, what changes and at what cost), and the audit became the second iteration. Crafting, companion and codex screens share one material language: soot, brass, worn paper, and type that could have been set in 1897.
Results
A complete UI kit across the core loops (dialogue, companions, crafting, codex, menus), with the HUD documented as a full research loop: first iteration, playtest, argued audit, rebuild, re-test. The re-test ran with a new group of twelve players on the same dusk-and-storm scenarios. The objective-recall probe went from seven of twelve to twelve of twelve, and time-to-orient dropped by roughly a third. Other parts of the build changed between rounds too, so I read this as a strong signal rather than proof that the HUD did it alone. (The figures are reconstructed for the NDA, but faithful to the pattern of the real findings.)
Key Takeaways
- Tone beats labels: Players read a reply from its colour, icon and wording, so I stopped adding labels.
- Review your own HUD: A ranked, argued audit was the cheapest way to redesign it.
- Quiet vitals: If you have to look straight at them to read them, they are already too loud.
Deliverables: Information Architecture, Dialogue System, HUD (2 iterations), Playtest & UX Audit, Crafting UI, UE5 Build (UMG, Blueprints)
Gallery
- Running in Unreal Engine 5
- The system map
- Dialogue: tone & consequence
- Narrative patterns: how the branches work
- HUD, first iteration
- The playtest: method
- The playtest: findings
- UX review: what works
- UX review: what to change
- HUD, second iteration
- Companions: trust as an interface
- Blacksmith: craft & upgrade
- Codex: the world on paper
- Title & menus (2 images)
- Wireframes (3 images)