Fluff Preset for DeepSeek R1: Click Here!
Read the information below, pretty please (￣▽￣)
PRESET-RELATED INFORMATION
Tested solely on "cute animalgirl who loves you" cards, but if you turn off/edit the story theme it should work fine on other types as well.
What's the difference between Concise and Draft thinking?
Draft makes all the thingies you don't really want to see in a response disappear. Well, kinda, and you have to edit it with your personal vices, but overall it trades ~20 seconds of generation (assuming unchanged response length and API speed) for a bit better response.
Other
- For R1 API, the prefilling seems to break the thinking box. Use this regex to hide it. There might be a better way of doing this but I know not of one.
- I'm using NoAss with User squash role, "User: " & "Model: " prefixes and checked "Preserve Assistant Last Message". How much it does I cannot say, but if you use it make sure to have the exact same settings.
CHANGELOG
2025/02/28 - changed how the thinking works.
2025/02/25 - updated the thing. It now follows the desired narration style better. Old Link: here.
2025/02/23 - made the thing. I don't think it will get any updates. Old Link: here.