FAQ

Why plain JSON files instead of a database?

Because everything narrative teams struggle with — versioning, merging, review, backup, tooling — is already solved for files in a repo. One entity per file means per-entity diffs; canonical serialization (sorted keys, stable formatting) means a one-line story edit is a one-line diff. Any text editor, script, or pipeline can read your story without Parlance in the room. More: git-native workflow.

Why is there no export step?

An exporter is a place where the authored truth and the shipped truth drift apart. Parlance's answer is a contract: the engine reads the exact files the editor writes, and the runtime semantics are published with conformance vectors that ports verify against. More: the engine contract.

Which engines work with it?

Any engine that can read JSON. The Godot runtime is public and conformance-verified; the TypeScript reference runtime that powers the editor's playtest is on npm as @orbitope/parlance-runtime from v0.16.0, MIT, so a web or JavaScript game can ship it as-is; the integration docs cover porting to anything else — the conformance suite tells you when your port is correct.

What's the licensing split?

Three layers, deliberately different:

The manuals on this site are © and not part of the MIT surface. More: licensing · the open spec.

How much does it cost?

Nothing. The editor is free — for hobby projects and commercial releases alike, with no seat count, revenue threshold, or license key.

Paid products come later and separately: hosted review links, hosted editing for teams, and a self-hosted server license — things that cost real money to run and are worth it to a team of ten writers. The local editor isn't one of them. More on that.

Is my story safe in AI training / telemetry?

Parlance is local-first: no cloud service, no accounts, no telemetry. Your files never leave your machine unless you push them somewhere. The optional AI drafting feature sends a request only when you press Draft or Regenerate, only to the provider you configure, with your API key — and it sends only the node you're drafting at (its speaker, its line and its existing choices) plus lists of up to 50 of your skill, flag, faction and character ids, so the model can reference only ids that exist. No other dialogue, lore or character description goes with it. Two other opt-in features use your providers the same way: the prose check's Sort these with AI sends the unknown words it found, each with a short sample of the line it came from, and placeholder voice sends a line's text to your text-to-speech provider when you generate it.

Does it handle localization and voice-over?

Yes — every player-facing string is extracted with a stable key; translators get flat JSON templates; coverage and stale keys (a key whose line was renamed or removed) are tracked per language, and VO maps voiceable lines — every spoken node line, never the player's choice text — to engine audio keys the same way. See the editor guide.

What are "dialogue offers"?

The mechanism that answers "which dialogue plays when I talk to this character right now?" — each dialogue declares an offer saying when it should play, and the engine picks the most specific eligible one. It's order-independent, with static checks for the classic mistakes, and it's the heart of state-aware conversations in Parlance: dialogue offers, explained.

Can a team use it without a server?

Yes — that's the design. Review (comments, suggestions, verdicts, narrative diffs) works with nothing but git: two people with plain clones have a full review workflow. See git-native workflow.

Something's broken. Where do I report it?

GitHub issues, with the version and steps — the form asks for what's needed. Parlance has no telemetry and no crash reporter, so a bug nobody files is a bug nobody knows about. Anything you can't say in public (licensing, security, a repro that would spoil your own game) goes to orbitopegames@gmail.com instead. The whole process, in one page.

Where do I start?

Play the demo (two minutes), then your first project (ten minutes), then how it all fits together.