Perchance AI roleplay: how it works, its limits and the alternatives

What Perchance is

Perchance describes itself as a platform for creating and sharing random generators. You write lists, lists pull items from other lists, and the result is a page anyone can open, edit and save as their own version.

The AI pages are generators like any other, built on Perchance's AI text plugin and its text-to-image plugin. According to the plugin's own page, text generation is too heavy to run on your device, so it runs on server graphics cards instead, and the bill is paid with ads that appear for visitors who are not logged in.

The pages you will actually use

• AI Character Chat: the fuller app, with threads, a character list, example characters, import and export, and image generation.
• AI Chat: a simpler page with example characters, boxes for the bot's description, your own description and the scenario and lore, plus an editable chat log.
• AI RPG: an adventure format, kept in the same kind of browser storage.

This article sticks to the first two, which are what most people mean by Perchance AI roleplay.

How a session works

1. Open AI Character Chat. The page loads with a starter character already greeting you.
2. Tap an example character to start a new thread, or build your own with the create button. A generate option can also draft a character from any web page you point it at.
3. Set your own name and picture, then type your first message.
4. Steer as you go: regenerate a reply, edit any message, write a line as the Narrator, or set a reply length in the character editor.
5. Add more characters with the pencil icon above the reply box to turn the thread into a group chat.
6. Export your data before you close the tab, for reasons covered below.

Writing a character for Perchance

The tips page gives advice that matches what works elsewhere. Example dialogue in the description or the opening messages is the strongest way to shape how a character talks. The description itself should ideally stay under about 500 words; anything longer goes in the lorebook, as short entries of about a sentence each that the AI searches when it needs them.

There is also a reminder note for writing instructions such as “be very descriptive”, though the tips warn that long reminders can confuse the model and suggest moving them to the top of the description instead. The {{user}} macro stands in for your name. If you already have a card from elsewhere, the import button accepts Tavern PNG cards, and the character card guide explains what each field is for.

Slash commands worth knowing

Typed into the reply box, these save a lot of clicking.

• /ai asks for a reply, and /ai followed by an instruction steers that one reply.
• /user has the AI write a turn for you.
• /nar posts a Narrator message following your instruction.
• /sum, /mem and /lore open the summary, memory and lore editors, and /lore followed by text adds an entry.
• /image draws the current scene, or whatever you describe after it.

How Perchance handles memory

Long chats get summaries. The AI Chat page and the AI RPG page both explain that special summary paragraphs are inserted once the text grows long, and that they extend the AI's memory. You may edit their content, but the pages ask you not to move or delete them.

The character chat adds memories and the lorebook on top, each with its own editor. Its tips mention that summaries and character memories can be switched off in the advanced settings for faster replies, at the cost of a less capable character. None of the pages checked states a context size in tokens.

The limits, as Perchance states them

• Storage is local. Chats and characters live in your browser, not on a server, so clearing site data deletes them, and the page warns that browsers sometimes clear storage on their own when space runs low.
• Backups are manual. The export button is the only safety net, and the tips note that export files can grow very large.
• Requests are rationed. The plugin page says each user can only have a few requests running at once, and the rest queue.
• Ads pay for it. Logged-out visitors see an ad on generators that use the AI plugin.
• Content follows the prompt. The plugin page says the model can produce adult material when prompted for it, asks people to prompt responsibly, and suggests adding an instruction if it drifts there unasked.

On privacy, the plugin page states that text prompts and responses are not stored on the server.

What could not be checked

The pages checked do not name the model behind the chat or its context window, and they mention no daily message cap. That does not prove there is none; it means none is written down where a visitor can see it. If either matters to you, test it in a throwaway thread before you invest in a long story.

Alternatives to Perchance

If browser-only storage worries you, or you want a context size you can actually look up, these are options whose facts were checked on their own pages. The roleplay chatbot comparison has the full table.

• SillyTavern also keeps everything on your machine, in a folder rather than browser storage, and lets you pick the model. It takes an install, and the setup guide walks through it.
• Janitor AI has a free house model with about 8k to 9k tokens of memory and a community character library, for users 18 and over.
• AI Dungeon suits players who prefer an adventure to a chat. Its free tier reads up to 4k tokens, and private single-player play is not moderated.
• NovelAI gives 50 free text generations to try its story editor, and encrypts stories in your browser before saving them on its servers.

Or start from a scene instead of a character: the AI roleplay chat here asks a few questions about your idea before you play.

Questions