⚗ The Lore Weaver's Cauldron AT THE TABLE
TURN—· DAY—
Your game
❓How to Playcommands & tips
Settings
🧠 Models
⚙ Advanced — Lorekeeper & background
Account
Leave
…
Lorekeeper processing…
0:00 / 0:00
▌
Enter send · Shift+Enter new line · /help commands · /guardian corrections · /undo undo · 🎤 speak (EN)
🛠️ Engine Room
0 lines
Waiting for logs… the DM's whispers will appear here.
📊 Usage
Samlar statistik…
🧠 About the Models
The Dungeon Master writes the world. The Lorekeeper quietly tracks HP, gold, XP and items in the background. The extraction model handles facts, the journal and summaries. You can switch the DM and Lorekeeper in Settings ⚙️ (the DM choice also appears when you start a new adventure).
qwen3.8-flash
The quick premium pick. Fast narration with solid reasoning — part of the 10€ unlock.⏱ fast · unlock · light thinking
deepseek-v4.1-flash
The quick thinker. DeepSeek's newest open-weight reasoning model — economical replies with real depth.⏱ fast · economical · reasons first
step-5-preview
StepFun's flagship — thinks deeply before every reply, so expect a longer pause. Part of the 10€ unlock.⏱ slow · unlock · deep thinking
step-3.7-flash
The steadfast default. Snappy narration, free for every account; occasionally needs a retry on complex character generation.⏱ fast · free tier · light thinking

Why does a "high reasoning" model take longer? Before answering, the model works through the scene — the campaign state, your choices, the rules — and you can watch it think in the reasoning bubble. More thinking means a richer, more consistent reply, but also a longer wait (sometimes 30–90 seconds). Fast models answer in seconds but with less depth. If a reply ever feels slow, that's the DM working — or switch to a faster model in Settings ⚙️.

👤
Adventurer
Loading…
🎨 Portrait
🎨 Avatar
click to enlarge
Generate a new avatar from your character sheet — items, traits and lore included. Or upload your own image.
🎨 YOUR OWN WORDS (optional)
Overrides the auto-built prompt from your sheet — describe exactly what you want generated.