Wyrdlyre

The lyre that plays fate.

A voice-driven AI Dungeon Master for tabletop role-playing. One player hosts, the rest join from their phones, and the story answers back in real time — with distinct NPC voices, atmospheric narration, and a deterministic dice engine you can actually trust.

What it does

• Listens to your party. Each player speaks into their own device. The DM hears you, transcribes, and responds.
• Narrates with mood. Pick a campaign tone — heroic, gothic horror, pulp adventure, gritty low-fantasy, or cartoon — and the DM adjusts its voice to match, scene by scene.
• Voices your NPCs. Each recurring character gets a distinct voice that persists across sessions, powered by OpenAI's expressive TTS (with optional ElevenLabs voice cloning on premium).
• Rolls real dice. A deterministic Python dice engine handles every roll — the LLM never fudges the numbers. Every result is audited and inspectable.
• Respects your privacy. Bring your own AI keys. Everything runs locally on the host machine — campaigns, keys, and audio never touch our servers.

How pricing works

Wyrdlyre is software, not an inference subscription. You pay a flat license for the app and use your own Anthropic, OpenAI, and (optionally) ElevenLabs accounts directly. No per-minute DM meter, no surprise bills — just the product and your own provider usage.

What ships at launch

• Dungeons & Dragons 5e via the Creative Commons SRD 5.1
• Party play across phones, tablets, and laptops, with live multi-device sync
• Voice in & out with five campaign mood presets and per-NPC voice routing
• Solo mode with AI party members who voice themselves and react in character
• Deterministic dice the AI never rolls — your trust intact
• Self-host option via Ollama for zero cloud-inference cost