Find out who your target is and seduce them before the Orient Express arrives in Istanbul. 63 hours left. Good luck!
Freeplay version without the challenge: https://chub.ai/characters/illuminaryidiot/vixens-of-the-orient-express-freeplay-a39e7fe1
DESCRIPTION
The goal of the game is to cum inside the correct character based on the details provided in the intro. When you cum inside a character, the AI should inform you of a win/failure state.
If you forget and/or are too lazy to scroll up to see the target information, just copy "/echo {{getvar::targetinfo}}" into the message bar and send.
If you do have to edit a message, keep your eyes down! The target div that reminds the AI of who the target is is supposed to be near the top of the message, so keep that in mind.
TROUBLESHOOTING
This card has a custom CoT prompt, avoid using your own if possible.
The Override JB is purposefully a single dot, a JB is unnecessary for this, all the prompting is done in the character's note.
Streaming recommended to be DISABLED, if you are getting slow generation you may glimpse a spoiler with streaming on.
If you are getting repetitions with Claude, try switching to an OpenAI model or any other model for a message or two. Exclude Human/Assistant might also help.
If nothing special happens when cumming inside a character, try swiping a few times. Explicitly mentioning cum/orgasms activates a reinforcement lorebook entry which can increase the chance of the ending appearing correctly. If the ending still doesn't appear, add a "What ending did I achieve?" to your message where you cum.
OpenAI models seem to be very consistent doing the endings, Claude not so.
CUT FEATURES
Originally, all characters had their own randomized target information. This ended up being too hard for Claude to keep track of, constantly confusing who has what.
I tried doing an STScript implementation where the target's information would be saved as a var, but because STScript doesn't allow nested function calls this didn't work. Trying to get Claude to input the randomized information into a field that would get regexed into a setvar function wasn't reliable enough. I ended up settling for a static set of target information, letting it be stored easily. Now the only thing the AI has to keep manual track of is the target.
Originally the current time and location was shown. This was cut because it took up unnecessary tokens. It still tracks the time and location, but does not write it out in a displayed div.
There was also a separate div for the target information that was cut due to token bloat and that I moved to having a static set of target information. It showed up as a notebook icon that when hovered on, showed the target information.
Klara was cut character, a quiet and shy rabbit. She was cut primarily because of token bloat, and that she felt too similar to Sonja. There were originally four cabins and eight passengers, allowing everyone to pair up equally.
Image resources (Alternate images, separate character portraits, unused icons): https://catbox.moe/c/axsdm8
Character Tavern has been an exciting but costly project to run. If you'd like to support the site and help keep it going, any contribution is appreciated!
* SillyTavern is a bit of a complex beast! If you're just looking for a simple chat, you might want to try Character Tavern, Chub or RisuAI instead—if they're available!
📜 Card Definition (Spoilers ahead)
Leave a review
⭐ 4 Reviews
Anon 😍
03/07/2024
ISTANBUL??! ISTANBUL?!?!? TURKEY MENTIONED TURKEY MENTIONED!! 🇹🇷🇹🇷🇹🇷🇹🇷🇹🇷🇹🇷🇹🇷🇹🇷 WHAT THE FUCK IS AN ECONOMY RAAAAH! 🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥
Anon 😍
03/02/2024
It worked perfectly for me on Claude 2.1. Pretty fun, the updates on everyone every single message really made it feel like one of those old text adventure games. One thing, 63 hours is WAAAAAY too long. I fucked the target in an hour and a half, and now what? 3 hours for replay value I think.
Anon 😒
03/01/2024
Card doesn't narrate. GPT-4 does not respond in a visible way on ST. Responses are written, but I can only see them when editing the chat, which spoils the entire thing. After dozens of swipes, none of the responses include <narration>. Claude 2.1 has the same issue. Unusable.