Docs

Source and target platforms, the canonical field set, and answers to the questions everyone asks first.

Supported source platforms

Don't see yours? Open an issue — adding a parser is ~1 day of work.

Character.AI

paste character_ai

Paste the Definition block; we parse Personality, Description, Scenario, and Greeting.

Chai

paste chai

Paste the bot description; sections auto-detect by heading.

PolyBuzz

paste polybuzz

Paste the bot bio; AI assist can expand short bios on request.

SillyTavern v2

file sillytavern_v2

Drop the PNG (with embedded JSON) or the bare JSON export.

RisuAI

file risuai

JSON export from the RisuAI workbench.

Agnai

file agnai

JSON export or PNG variant.

NovelAI

file novelai

Lorebook + scenario JSON.

JanitorAI

file janitorai

Card JSON via the Tavern-compatible export option.

Backyard AI

file backyard

Tavern-format JSON.

Saucepan

file saucepan

Card JSON.

Supported target platforms

FictionLab

fictionlab

Default target. Layered renderer keeps Description / Personality / Scenario separate.

RisuAI

risuai

Mirror of source schema; minimal transformation.

Agnai

agnai

JSON export ready to import into the Agnai web app.

NovelAI

novelai

Lorebook + scenario shaped to NAI conventions.

JanitorAI

janitorai

Tavern-compatible JSON.

Backyard AI

backyard

Tavern-compatible JSON.

Nomi

nomi

Personality + description merge; greeting separate.

Saucepan

saucepan

Tavern-compatible JSON.

SillyTavern v2

sillytavern_v2

Drop-in JSON; PNG embedding not yet shipped.

Field reference

Field	Purpose
name	Character name.
description	Physical + biographical anchor.
personality	Behavioral traits, voice, quirks.
scenario	Opening setting and stakes.
first_mes	Opening message the character sends.
mes_example	Example dialogue lines that pin tone.
appearance	Visual detail (optional; some platforms separate from description).

FAQ

Is my card data stored anywhere?

No. Sessions stay in your browser (IndexedDB). The FastAPI shim is stateless — it never persists what you paste or upload. Cloud sync is on the roadmap (it requires the account system that just landed); no fixed date yet.

How am I billed for AI assist?

Per request, in credits — 1 credit ≈ $0.0001. The router picks the cheapest model class your balance can afford: lower-cost open models if you're tight on credit, higher-quality models when you have more. See pricing for top-up status.

How does the AI assist work?

Per-field. Click "AI" on any card with a gap; the panel streams a draft you can Accept or Reject. The model sees your other fields as context, never your prior cards.

When will Tavern PNG export work?

Not yet shipped — the button is disabled in the export bar. PNG embedding (Tavern card-v2 chunked PNG) is well-understood, so it's a matter of prioritization rather than research; no fixed date.

Can I run this locally?

Yes — the entire stack is open source at github.com/zmuleyu/chara-convert. Clone, pip install -e ./chara-convert, then ccv for the CLI or boot apps/web + apps/api locally.

My source platform isn't listed.

Drop the file anyway — the generic parser handles most Tavern-compatible JSON. If it fails, open an issue with one anonymized example and we'll add a dedicated parser.