YanMateYanMate
How does your character actually talk?
Paste a character's replies, or a whole conversation with them, and this page reads the voice out of it: how long the sentences run, how often they ask or exclaim or trail off, the words and phrases that keep coming back, and formatting habits like narrating actions in asterisks. It turns those into plain notes you can put in a persona, and picks the replies that show the habits best as a ready example-dialogue block for a character card. Everything is counted in your browser — nothing is uploaded, and no AI rewrites a word.
Read a voice
Paste the character's replies with a blank line between them, or a whole conversation in Name: message lines — the transcript the Character.AI export viewer saves works as it is. Five replies is a start; twenty is better.
Read in your browser. Nothing you paste is uploaded or saved.
Why example dialogue beats description
A persona that says a character is “dry, clipped and sardonic” asks the model to invent what dry and clipped sound like, and every model invents something slightly different. Three replies that are dry and clipped leave nothing to invent. Models imitate examples far more faithfully than they follow adjectives, which is why the example-dialogue field is where a voice is really set.
The catch is that good examples are hard to write from nothing, and people who have a character they already love usually have something better than a blank page: hundreds of replies showing exactly how it spoke. The hard part is reading them for habits rather than story. That is the part this page counts.
From a transcript to a card
1. Gather real replies
A transcript of a conversation you liked is the best source there is, because it shows how the character spoke rather than how someone described them. If you are moving a character from Character.AI, the export viewer turns the archive they email you into plain-text transcripts this page reads directly. Otherwise copy a chat out of the app you used, or paste the replies you already picked with a blank line between each.
2. Read the habits, not the plot
The notes list what is unusual about the voice — a character whose sentences average five words, who opens every other reply with "Well", who asks a question in one sentence of three. These are what an adjective in a persona fails to convey. What happened in the conversation does not matter here, and the numbers ignore it.
3. Put the examples in the card
The example dialogue is made of the replies that show the most habits at about the character's usual length, with the line each one answered kept above it and names swapped for {{user}} and {{char}}. Trim any that repeat a plot point you do not want the character bringing up, then paste the block into the card's example-dialogue field — or open it straight in the card maker.
4. Test against the old replies
Talk to the rebuilt character using prompts you used before and compare its replies with the transcript. If it drifts, change one field at a time: a note in the persona for a habit that is missing, a different example for one it is overdoing. Close is a realistic target; identical is not, because a different model reads the same card in its own way.
The whole move, from requesting your data to testing the rebuilt character, is in the guide to moving your characters from Character.AI. For the persona and greeting around the examples, how to write an AI character covers the rest of the card.
What counting cannot see
Everything here is counted, not understood. It can tell you a character asks a question in one sentence of three; it cannot tell you whether those questions are tender or interrogating, or that the character is grieving. Tone, mood and what a character cares about are still yours to write — the notes are the part a writer tends not to notice, not the whole of a voice.
The numbers also need enough text to mean anything. A habit is something that comes back, so three replies show formatting and not much else, while twenty replies from more than one conversation start to show the voice rather than one scene. Sentence splitting, the common words it ignores and the contraction count assume English; formatting habits and the example picker work in any language.
Example dialogue costs context like anything else on a card. The block here is kept to about 600 tokens, and the context window simulator shows when a frontend will drop it to make room for the conversation.
Questions
- What is example dialogue on a character card?
- It is the mes_example field: a few short exchanges, each starting with <START>, that show a model how the character talks. Lines beginning {{user}}: are the other person's and lines beginning {{char}}: are the character's. A model imitates an example far more reliably than it follows an adjective — "speaks tersely" in a persona leaves it to guess, while three terse replies show it. Most frontends send the examples early in a chat and drop them as the conversation fills the context window.
- How do I make an AI character sound like a character I already know?
- Start from how they actually spoke, not from a description of them. Paste a transcript of a conversation you liked into this page, pick the character as the voice to read, and it lists the habits that stand out — sentence length, questions, openers, repeated phrases, narration — and picks the replies that show those habits best as example dialogue. Put the notes in the persona and the examples in the example-dialogue field, then test against the old replies and adjust one field at a time.
- How many replies does the analyzer need?
- It will read as few as three replies and eighty words, but that is only enough to see formatting habits. Twenty replies or more, from more than one conversation, is where sentence lengths, openers and repeated phrases become trustworthy, because a habit is something that comes back rather than something that happened once. A whole exported conversation is ideal.
- Does it use AI to write the example dialogue?
- No. Nothing on this page calls a model. The analysis is text statistics counted in your browser, and the example dialogue is made of real replies from what you pasted — chosen for showing the habits it found and laid out in the card format, never rewritten. That keeps the examples in the character's own words, which is the point of them.
- Does it work for languages other than English?
- Partly. Formatting habits — asterisk narration, ellipses, dashes, emoji, line breaks — are read the same in any language, and the example-dialogue picker works on any text. Sentence splitting, the list of common words it ignores and the contraction count all assume English, so for other languages those numbers are rougher, and the page says so when most of the text is outside the Latin alphabet.
- Is my transcript uploaded or stored?
- No. The text is read by JavaScript in your browser and never sent anywhere, and unlike this site's other tools nothing about it goes into the page address, because a conversation does not belong in a link. The fingerprint image is drawn in your browser too, and carries only the numbers, the name you gave and one repeated phrase. If you open the example dialogue in the card maker, it is kept in your browser's storage as part of that draft, and nowhere else.
Other free tools
- Character card viewerDrop a character card — PNG, JSON or .charx — and read every field, warning and hidden value inside it, then convert between V2, V3 and PNG. Runs in the browser; nothing is uploaded.
- Character.AI export viewerOpen the ZIP Character.AI emails you and read it: your characters as cards you can take anywhere, your chat history as plain text. Runs in the browser; the archive is never uploaded.
- Character card token counterDrop a card or paste a persona and see the token cost field by field — separating what is re-sent on every single message from what is sent once or only sometimes. Runs in the browser.
- Context window simulatorSet a context window and a character card, drag the conversation forward, and see how many exchanges the character still remembers — and at which message it starts to forget. Runs in the browser.
- Character card makerFill in a form and download a real character card PNG carrying both the V2 and V3 chunks, so it opens anywhere. Any image works — a JPEG is converted in the browser. Nothing is uploaded.
- Character card templatesEight original V3 starters — mentor, rival, cozy companion, game master, language partner, study coach, noir detective, historical figure. Download PNG or JSON, or open one in the maker. Nothing is uploaded.
- AI companion cost calculatorPick how long you would keep an AI companion app and see what each plan costs over exactly that stretch, monthly against prepaid annual — every price sourced and dated, blanks where none is published. Runs in the browser.
- Character safety checkPaste a character's persona, its greeting or a reply it gave you and check it against a published list of 30 guilt, dependency, exclusivity and minor-coded language patterns. Every finding names the rule it broke. Runs in the browser.
An AI character on YanMate.