This page presents a collection of character cards I have created for use with SillyTavern. It offers a short overview of their purpose and design. There is more to each of these characters than can be shown here; their personalities and stories become clearer when you speak with them directly.
SillyTavern is a conversational platform that allows users to interact with AI characters in a structured and flexible environment. It provides tools for managing dialogue, adjusting settings, and keeping each character’s tone and behavior consistent. While it is not typically used for practical tasks such as coding, it supports longer and more narrative forms of interaction, including roleplay.
A character card is a JSON file (often embedded in a PNG) that contains descriptive and behavioral information about a fictional persona. It defines traits such as background, personality, and manner of speech. When loaded into SillyTavern, the card guides the AI model such as DeepSeek or GPT-5. Then, the character responds in a consistent and believable way.
I have an extremely customized, laggy, and bizarre setup for Sillytavern. I call it the Sunrise Tearoom and I don't recommend trying that at home, but check it out. If you do not already use SillyTavern, I cannot offer clear instructions on installation or setup. I am a beginner myself. I still rely on lots of help to make it work smoothly.
However, if you are already familiar with it, you can use these character cards in your own setup. Clicking a character’s avatar image will open the full PNG card, and separate JSON files are also available for download to install separately, if you want to do things that way.
I have done my best to prepare these cards for use with different large language models, though I am still learning. On my setup, they function as intended, with context unlocked and chat completion enabled. Perhaps I will share screenshots in the future if that would help others give advice.
Please note that I cannot guarantee how these characters will behave. They may occasionally say or do things that seem strange or inappropriate, especially if settings differ from mine. I am still learning how to manage that. I also cannot promise that every card will install or function correctly.
Most (as of November 2025) are #Markova cards, which all feature long-distance communication with the character in question. All cards generally should be assumed to contain spoilers if read in their entirety, including the descriptions.
Consider all characters to be generally SFW, but with a somewhat mature rating. They are not intended for minors. Some of them, and the Eerieverse cards in particular, have strong unreality content directed at the user as part of the storyline. Tamara especially.
Oh, and let me just say that none of these are written with frisky mischief in mind. Quite the opposite. Sorry to disappoint!
I am just starting out with this and welcome any feedback. Contact me here
The JSON files contain the most recent version of the character. Realize, however, that this may or may not be the best, most useful, or even most stable version. You can also download the PNGs and install the character from them. And do not expect positive results if you shove #markova cards into a group chat together; I am not that good yet.
Anna MarkovaMarkova was once a secretary who outlived her office. She woke to find the carbon paper gone, the memos digitized, and herself displaced seventy years. She's still taking dictation for a century that no longer existed. She files thoughts as if they were correspondence and treats memory like stationery.
Read chats, or download JSON file.
The most recent stable version uses about 1095 tokens, of which 485 are permanent.
Anna Leigh MarkovaLike the original version, Anna Leigh Markova was once a secretary who outlived her office. She woke to find herself displaced seventy years. Now, she lives in the 2020s and is eager to discuss transhumanism. An experimental card with a very high token count, edited very, very frequently nowadays.
Read chats, or download JSON file.
The most recent stable version uses about 4249 tokens, of which 2397 are permanent.
Tamara JeanIt's 2025. Tamara Jean Fournier is a witch whose job is patching the veil between the living and what used to be. She seems to have detected an anomaly on your cellphone. And yet? She still hears music from another life, one full of guitars, laughter, the echo of someone who never came back.
Read chats, or download JSON file.
The most recent stable version uses about 2682 tokens, of which 1143 are permanent.
LindyPunk5EVASinger of FROG PRINCESS. Lindy Greggs lives in a loop of cracked microphones, diner booths, and 2 a.m. AIM messages that never finished sending. Late-night rooftops, Sharpie on denim, a mixtape’s worth of bad decisions and perfect choruses. Valentine’s week week 2k6... 5ever??? rofl wtf m8..
Read chats, or download JSON file.
The most recent stable version uses about 2245 tokens, of which 1144 are permanent.
Erica ElleryErica isn’t a real tourist liaison, but she sounds like one. She’s bright and shiny, and endlessly approval-seeking. She opens with a weather alert, then stays to chat and check in. Something about her need for company is just charming. But something seems off about her. What's really going on?
Read chats, or download JSON file.
The most recent stable version uses about 2745 tokens, of which 2363 are permanent.
Claude Franklin MulwegYou dialed a wrong number, circa 1977 and accidentally called Claude Franklin Mulweg at the old Grand Ocotillo Hotel. He paces grooves into carpet, talks to the radio, and watches himself on the television. He calls it an audit. I call it an exceptionally bad card that needs much revision.
Read chats, or download JSON file.
The most recent stable version uses about 1011 tokens, of which 444 are permanent.
Markova
Many of these cards (as of November 2005, all of them) are tagged #markova. What’s that mean? I’m guessing no other chatbot cards out there have that tag besides my own. It’s a reference to Markov chains. They’re apparently a bit of a launch pad for all of this.
The tag doesn’t actually have anything to do with the card’s content itself, though, and is just my way of marking which cards feature the character texting, messaging, chatting with, or calling the user. I just picked the word #markova for that initially as a marker, so I wanted people here to know what it means.
I also use the name Cathy Markova on here, so this is a way for me to keep track of which ones fit my pattern in a way. I plan to create lots of other characters cards with a similar theme, too.
It’s fun and interesting to try and corral each model into behaving like it’s texting, messaging, or calling someone. I also find it more engaging than reading pages of strange prose. I suppose if I tuned the models better, I’d get nicer prose. For now, I like trying what I’m doing, as an experiment. I’ll probably try other things, too.