Features
Everything in the Rin Chat desktop app
Rin Chat covers the whole loop: the character card and the library around it, the chat and its cast, the prompt that reaches the model, and the encrypted store underneath.
Character cards
Bring in the cards you already have, edit every field, and keep a large library in order.
Import existing cards
Rin Chat reads common character card formats, including cards embedded in a PNG image. You can import a whole folder at once.
Structured editor
Each part of a card gets its own field: description, appearance, personality, behavior rules, speech style, background, setting, greetings and example dialogue. Older cards migrate to that shape when they load.
Library
Folders, tags, favorites, and sorting by recently updated, recently chatted, newest or name. Search covers the text inside a card, and inside a character it also searches the messages of every chat with them.
Tag manager
Rename, merge, color and delete tags across the whole library, normalize their capitalization, and review suggested merges for near-duplicate spellings.
Version history
The editor keeps snapshots of a card as you leave it. Open the history and restore an earlier version.
Checks while you write
A live token estimate sits next to each long field, with advisory notes such as a missing first message, an oversized description or a lorebook entry with no keys.
Chat
A writing surface built for iteration, not a single take.
Streaming replies
Text arrives token by token, so you can read a reply as it is written.
Swipes and regenerate
Ask for another reply, or steer the retry with one-click dials for length, pace, tone and repetition. You can also write your own direction for the retry.
Edit, delete and branch
Change any message in place, remove it, or branch the chat from a message to explore a different direction while the original stays intact.
Continue and impersonate
Extend the last reply with more text, or have the model draft your own next line into the input box, where you read and edit it before it sends.
Response length and prefill
Cap the next reply at short, medium or long without touching your sampler settings, and start a reply with exact text you supply.
Quick replies
One-tap buttons above the composer send a message you saved, or drop it into the input for you to edit first. They support macros.
Export and import
Move a chat out as Markdown, plain text or JSONL, and bring it back in.
Casts and personas
Decide who is in the scene, including you.
Group chats
Put several characters in one chat. Turns rotate, and the chat resumes from the last speaker when you reopen it.
Per-member settings
Each member of a group can use its own provider, model and prompt preset.
The model picks the speaker
Ask who should speak next. The model reads who was addressed and what fits, and a plain rotation takes over when that is unclear.
Auto mode
Let the members answer each other for a set number of turns while you read.
Personas
A persona holds your display name, the name the model sees, your appearance, your personality and example lines of your dialogue.
Temporary personas
Write a persona that belongs to one chat only, then promote it to your saved list if you want to keep it.
World building
Background detail that reaches the model when it is relevant, and not before.
Lorebook editor
Full control of a card's entries: keys and regex keys, secondary keys and their logic, always-on entries, insertion position and order, depth and role, and a probability.
Standalone worldbooks
Keep a worldbook outside any single card, in its own editor. Turn one on for everything or attach it to a single chat. World info JSON imports and exports.
Lorebook tester
Paste sample text and see which entries would activate, and why each one matched, missed or was filtered out.
Macros
Insert dynamic values into a card, a lorebook entry or a prompt: variables, comparisons, conditionals, dice rolls, random picks, date and time. A settings panel resolves any macro against a sample scope so you can test it first.
Generators
Reusable weighted randomizers for NPC details, dice and outcome tables. Call one inline and the result freezes into the message, give it a chance to fire on its own each turn, or let the model call it.
Prompting and memory
The controls that decide what the model receives, and what it still knows fifty messages later.
System prompt presets
Save system prompts and switch between them, along with the context template that assembles the character block.
Author's Note
Hold a standing instruction near the end of the prompt, where the model weighs it most. Set a default, override it per chat, and choose its depth and role.
Context budgeting
History is trimmed to the token budget you set. The system block and the newest message are always kept.
Summary and memory
Condense earlier turns into a running summary, on demand or automatically. Pin any message so its content is injected on every turn.
Scene tracker
A tracked note of the current scene — place, time, who is present, the mood — injected as its own block so those details stay consistent.
Regex scripts
Ordered find-and-replace passes over message text. Choose whether each one affects the prompt, the display or both, and try it on sample text first.
Models and providers
Bring your own endpoint, and control how it decodes.
Any OpenAI-compatible provider
Add as many providers as you need, each with its own endpoint, key and model. The key stays on your device.
Model picker
Type a model id directly, or search the list the provider reports and pick from it.
Sampler settings
Temperature, top-p, top-k, min-p, penalties, seed, stop sequences and a response token cap, plus advanced samplers including DRY, XTC, mirostat, typical-p and CFG.
Reasoning controls
For models with a reasoning step: the effort level, the request format the provider expects, and whether the reasoning is shown. The reasoning is never sent back to the model.
Fallback and retry
If the main model is overloaded before anything streams, Rin Chat retries once on a fallback model you named, and offers a model picker on the error itself.
Optional subscription
A subscription covers the AI usage through our proxy, so there is no key or endpoint to configure at all.
AI helpers
Optional assists that run on the provider you already selected.
Suggested replies
Get candidate replies for yourself when you do not know what to write next.
Rewrite a section
Highlight part of any message, add an instruction, and have only that part rewritten with the rest of the chat as context.
Chat titles
Name a chat from its opening exchange, on request or automatically.
Story director
Give it a storyline. A planner turns it into an outline you edit, regenerate or reject, then a second pass writes the direction for each beat. Run it beat by beat with a review step, or let it play with a pause for reading.
Card tools
Summarize a card into a private note, rewrite a single field from an instruction, and restructure a flat imported card into the structured fields.
Editable prompts
The quick prompts behind the AI edit buttons and the director's own planner prompts are yours to change.
The app itself
Desktop software, and the settings that make it comfortable to read and write in.
Dark interface
One dark theme, matched to aicharactercards.com. Adjust the interface scale and the message font size from inside the chat.
Formatting shortcuts
Select text and press a shortcut to wrap it in italics, bold, quotes or code, or right-click for the same menu. The shortcuts are yours to redefine.
Readable messages
Messages render as Markdown, with spoken lines in quotation marks colored so dialogue stands out from prose.
Guided tour
A one-time tour points out the chat controls on first run, and you can replay it at any time.
Prompt inspector
See the exact request and response for each call, the token budget per section, the lorebook entries that fired, your variables, and the generator log.
Statistics
Per-chat counts of turns, words and estimated tokens, plus a library view of totals, favorites, unfiled cards, cards never chatted with and top tags.
Automatic updates
New versions install from inside the app. There is nothing to re-download by hand.
Privacy and your data
Your library stays on your machine, encrypted, and you hold the keys to it.
AES-256-GCM encryption
Cards, chats, images and settings are written to disk as encrypted blobs.
Encrypted index
The database holding names, titles, tags and folders is encrypted as well, so a file listing gives nothing away.
Unlock on your device
A key held in your operating system keychain unlocks the app on your machine. Windows Hello or Touch ID can gate that unlock.
Recovery kit
A 24-word recovery phrase, kept as a file or written down, is the way back in if the machine or its keychain is lost.
Backups
Export a full encrypted backup and restore it on another machine, or keep rolling local snapshots in a folder you choose.
Separate vaults
One computer can hold several independent vaults, each with its own key, data folder and library.