⚠️ Storage unavailable — this session lives in memory only and vanishes on close.
🎲
Tale Engine
Private preview — enter the passphrase.
🎲
Weaving the world…
🎲
--
Rolling…
💀 THE END
⚙️ Engine Settings
Your key vault. Add any keys you like — each new adventure picks which engines to use, so you can run different models for different stories. Keys never leave this browser.
Many models, one key — openrouter.ai/keys
Claude direct — console.anthropic.com
GPT stories and gpt-image-1 scenes — platform.openai.com
Gemini stories — aistudio.google.com
Kimi stories — platform.moonshot.ai
DeepSeek stories — platform.deepseek.com
Ollama, LM Studio, Groq, Together… A local server is the most private option of all.
FLUX scene images — fal.ai. Validated on your first scene.
Change the story model, switch the image model, or turn scene images on/off — mid-story, applies from the next turn.For bug reports — errors & device info only, never your keys or story.
Fine-grained token for the project repo. Only ever sent to api.github.com.
Commits the full playthrough — actions, rolls, GM outputs & secrets, timings, scene images — to the repo's playtests/reviews/ for critique.
Private by design: keys & saves never leave this browser.
✨ New Adventure
⚠️ A story is in progress — ⚔️ Begin starts a new one. ▶ applies your tweaks here (engines, images, style, rating, pacing) to the current story and resumes it; premise, genre & theme edits only shape a new story.
First time here? Add your API keys, then you'll land right back in this wizard.
🎲 picks from the built-in bank instantly · ✨ asks your story engine to invent a fresh premise for your selected genres/themes.
Sets story tone and image safety level. Your providers' own policies always remain the ceiling — this app never bypasses them.
Guides the game master's pacing — how fast arcs escalate and resolve.
How much story each turn delivers. Brief for a brisk pace, Rich to sink in.
Schnell is the sweet spot for speed and cost. For realistic/photographic styles pick FLUX dev/pro, HiDream, or SD 3.5 — schnell's 4 steps produce anatomy glitches (extra limbs, warped faces). Each label notes the model's strength. Custom accepts ANY fal image endpoint id — browse fal.ai/models and paste (image models only). Any SD/SDXL checkpoint runs community models of every style — photoreal, painterly, anime — through fal. Checkpoint field: for a single-file model paste the DIRECT file link — HuggingFace's download/"resolve" link or a Civitai download URL (HF "blob" page links and "owner/repo/file.safetensors" paths are auto-converted); a plain "owner/repo" id works only for diffusers-format repos. Your Art & Narrative Style still sets the look; use Custom style for checkpoint-specific tags.
A frontier model here designs the world's secrets, twists, and arc revisions — a handful of calls per campaign, pennies — while the story engine above narrates every turn. Leave on "Same as story engine" to use one model for everything.
📖 Journal
💾 Save & Load
Done with this story? Exit shelves it in a free 💾 slot and returns to the home screen.