I wouldn’t be here without guides from the community, so I make this in hope of helping others who need it.
Disclaimer: This is what personally works for me. It may not work for you, or against your taste. Please read and use of your own accord and responsibility.
This was made by: MAND
MAND's Personal Settings
Advanced prompt:
For JLLM: kolach3
Below are prompts for proxy:
Currently using:
No additional plugin or nsfw prompt required.
By @fibal81080
Check out his direct link: https://rentry.co/MLRPE
Some may say it’s too long, but it works nicely and covers everything and every genre you could ask for in a role-play. It’s best to use it in platforms that charge per message or free models.
By Brbiekiss (she already closed her doc? Idk)
Check out her direct doc: https://docs.google.com/document/d/1xfJOM3KnxKqQLD_I29nXGj2u-gz_RAs8ojqtXy3sSQE/edit?usp=sharing
She gives lots of better guides and tips than me.
For NSFW:
For more toxic/ dead dove characters:
Images
On Tensor:
- Flat color style:
- Model: Zephyr Flat Color
- Lora: Detailed Painted Semi-Realistic Style at 0.8
- Prompt:
masterpiece, best quality, absurdres, newest, very aesthetic, highres, (tsurime:1.1), (flat color:1.5)
- Semi-realism style:
- Model: AVALANCHE - Semi Realistic Illustrious
- Lora: Meteor - Illustrated HandDraw at 0.8 + Crayon Semi Realism || Nijijourney at 0.5
- Prompt:
beautiful lighting, perfect contrast, detailed, clear lines, 4K, pastel color, portrait.
- Chibi style: Niime - Illustrious, Aphantasia Illustrious, or FlatNiji Illustrious + Lora: USNR STYL, chibi sticker style (With prompt "dynamic pose", it can generate a whole sticker pack, great!)
- Anime style: ToxicEchoIL + Lora: USNR STYL (vibrant coloring), atmosphere IL, Dark Darker Darkness (dark lightning). Dark Aesthetic || Nijijourney (semi-realism lora to make anime style more adult-ish)
- Model: Illustrious - MultiStyle Variation A
- Lora: Meteor - Illustrated HandDraw at 0.8 + Anime Enhancer XL at 0.8
I edit images on Photoroom and Canvas.
On Niji Journey with VIP personalization + prompts: semi-realism, high quality, dynamic pose. Due to being broke I don't use this anymore.
Chat Memory Prompt
I use this for proxies after >100 messages. It works wonders.
However, there are redundant commands in this memory so you gotta read and take out what’s not necessary.
MAND's Tips
Guide for people that don't know what proxy is, start with deepseek: Click here
Sophia Lorebary is a third-party service that provides amazing plugins and a lorebook system to enhance your role-play. It has its own community developing open plugins and lorebooks, and you can participate to create your own. Instead of the provider's URL, you use hers. Check out her site for more info.
Why I don't use Chutes (long rant. This doesn't mean I'm discouraging you from using it. Make your own decision): I had first hand experience how messy Chutes was. They used to provide 200 msg/ day for free, then they made up lies about people making 10000 accounts to bypass the system and punished everyone, changing it to a permanent $5 for 200 msg/ day + 50 mgs/day for free user. But that wasn't enough for them. They said they would prioritize users directly from them, leading to constant error 429 on OR. Now it's $3/ month for 300 msg/day. What a shitty, greedy company. It all happened in the span of one to three months. They got greedy so fast. Therefore, I won't put a dime into Chutes!
LLM Review
I use on OpenRouter so these models can be found there. It's ranked from best to worst in my opinion.
Proxy URL: https://openrouter.ai/api/v1/chat/completions
My criteria: Long response, can pick up subtle hints, has nuanced understanding of both {{char}} and {{user}}'s persona, and can continue the story naturally, interestingly
Free models
z-ai/glm-4.5-air:free : It gives long reply and continues the story naturally, good with sex scene. Personally I love it, especially on lazy day when I only need to write 1 sentence and let it do the heavy lifting. Weakness: doesn't pick up nuances or small details well, can be OOC, suitable for story-driven role-play that don't have complex characters.
minimax/minimax-m2:free : Is faster than glm and gives long detailed reply. However is censored model so harder to get nsfw response from it even with PREFILL. Has problem with understanding lorebook too, works weird when there's a lore entry. Other than that it's good.
deepseek/deepseek-chat-v3.1:free: This is the standard and most famous LLM for role-play right now. With the prompt above, everything is in "standard" mode: decent wait time, decent response, and it sticks to the personality and prompt while still subtly developing the story. Sadly resources are overwhelmed all the time.
deepseek/deepseek-r1-0528:free : This was the best model for role-play before Chutes became unreliable and blocked usage from OpenRouter, which is supposedly their own provider. Great for comedy and chaotic vibes. It's still good if you don't mind some 429 errors.
cognitivecomputations/dolphin-mistral-24b-venice-edition:free : bot gets a bit too talkative and OOC.
meta-llama/llama-3.3-8b-instruct:free : It's as good (or bad) as Deepseek V3.1. I felt no difference. It needs a lot of handholding and won't move the story forward unless you explicitly tell it to.
cognitivecomputations/dolphin3.0-r1-mistral-24b:free : Replies are long and detailed, but repetitive, going in circles.
openai/gpt-oss-20b:free : Weird case. Sometimes it gives the best response, sometimes it's just gibberish, nonsense that repeats itself. It follows the prompt well in first few messages, then falls off the more I go. It has censorship.
mistralai/devstral-small-2505:free : Generates gibberish. Don't use.
"Speaks for You" group:
mistralai/mistral-small-3.2-24b-instruct:free: Super fast and good. Only bad thing is it speaks for you, but it lowkey cooks. Its speed makes up for everything.
qwen/qwq-32b:free : It writes a whole novel with beginning and ending in 1 reply. But someone may be into it.
qwen/qwen2.5-vl-32b-instruct:free : Plain and boring.
Paid models
Price from high to low: claude - gemini - aion - mistral - deepseek.
anthropic/claude-sonnet-4: Divine. A game-changer. It's very nuanced, but it requires a well-written bot to maximize its abilities and your money's worth. Very expensive.
aion-labs/aion-1.0 : Its response is long, very nuanced, and naturally continues the story in the best way possible. It's desgined for role-play specifically. Worth trying.
deepseek/deepseek-v3.2-exp: Contrary to others who say Deepseek fell off with its new version, I think deepseek-reasoner is better than R1 0528. It listens to the prompt and OOC commands so much better and can be subtly nuanced in its response. However, it's not cheap. The direct URL is: https://api.deepseek.com/chat/completions
google/gemini-2.5-pro : It's somehow fallen off. It used to be so good, but now even with the paid version on OpenRouter, it's just slightly better than Deepseek V3.1, and I feel like that's not worth the price. I still use it occassionally though. You can also use gemini from Sophia Lorebary's url but Google has strict censorship rules and your API may get banned. All of my APIs got banned. :(
mistralai/mistral-medium-3.1 : It's also slightly better than deepseek, but more expensive so I leave it below deepseek.
How I use the features
- Example Dialouge to set dynamic among {{char}}, {{user}} and other NPCs;
- Lorebook for secrets, hidden events and easter eggs.
More details below:
Example dialogue
I've found that example dialogue functions like a previous chat log for a LLM, establishing canonical past events for the user's character. More advanced models will replicate exactly what's written there if they encounter the same scenario or trigger, which makes the approach ideal for people who want both their character and other NPCs to behave more consistently and rigidly. As a result, the models' responses are less random or creative.
Example dialogue doesn't consume permanent tokens, but it does fill up context tokens. This makes it exclusively suitable for models with a large context window, such as proxies, unless a larger context length becomes standard for JLLM in the future.
Scripts/ Lorebook
Scripts are best for fantasy or game-type bots, hidden events, or secrets. I use them for specific factions (like a well-known clan, family, or company), hidden NPCs, or secret places in the world.
A script won't trigger unless the user says an exact keyword. Therefore, if you have worldbuilding, NPCs, or traits that the bot should automatically use, you shouldn't put it in the scripts.
If you're confused, I recommend not using scripts. A good combination of Personality, Scenario, and Example Dialogue is more than enough to create a lively and responsive bot. There's no need for scripts unless you have extremely heavy lore or a complex system you want to implement.
Other useful resources (for decoration, maybe for my future use, or not): https://rentry.co/freyatut
My profile CSS
This was made by guides written in my description and chatgpt. I have no knowledge of CSS so it may need some clean up.
This is the content in profile box.
Putting the content you want in specific class helps you customize it in style.
This is the customization of everything.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 | <style>
@keyframes flicker {
0% { opacity: 1; }
49% { opacity: 0.95; }
50% { opacity: 0.7; }
100% { opacity: 1; }
}
@keyframes hazyMistGlow {
0% {
opacity: 0.5;
filter: blur(4px) brightness(1.2);
}
50% {
opacity: 0.8;
filter: blur(8px) brightness(1.5);
}
100% {
opacity: 0.5;
filter: blur(4px) brightness(1.2);
}
}
@keyframes shimmer {
0%, 100% { opacity: 1; }
50% { opacity: 0.7; }
}
@keyframes scanlines-glitch {
0% { background-position: 0 0; }
100% { background-position: 0 100%; }
}
/* Info Box Content Customization */
.Series {
display: block; text-align: center; color: rgb(222,94,106); font-size: 24px; text-decoration: underline; margin: 0 auto; width: fit-content; font-family: 'Pirata One', cursive;
}
.Series:hover {color: #5eded2}
.Links {
color: rgb(222,94,106); text-decoration: none;
}
.Links:hover {color: #5eded2}
summary {
cursor: pointer;
outline: none;
list-style: none;
text-align: center;
width: 100%;
font-weight: 700;
font-size: 1.8em;
padding: 5px;
background: #080603;
color: #5eded2;
font-family: 'Pirata One', cursive;
letter-spacing: 2px;
border-radius: 20px;
margin-bottom: 0.5rem;
margin-top: 0.5rem;
border: 3px solid #5eded2;
}
details[open] summary {
background-color: #080603;
border: none;
}
details p{
padding-left: 1rem;
padding-right: 1rem;
padding-bottom: 0.5rem;
}
details[open] {
background-color: #080603;
border: 3px solid #5eded2;
box-shadow:
0 0 8px #5eded2,
0 0 20px #5eded2;
border-radius: 20px;
max-height: 400px;
height: 100%;
overflow: hidden auto;
}
/* Info Box Appearance */
.css-1bn1yyx {
width: 90%;
max-width: 420px;
height: auto;
margin: 20px;
display: flex;
list-style-type: none;
flex-direction: column;
flex-wrap: wrap;
cursor: pointer;
}
.css-8g8ihq {
align-items: center;
}
/* Custom Font */
h1, h2, h3, h4, .css-1cd408u, .css-p641ht, .css-1qj3ulc, .css-zgqw37, .css-w86fka,
.css-1uodvt1, n.css-hjkukh, .css-1x04uxb, .css-kzd6o0, .css-1jj3srb, .css-6su6fj, .Btn2-purple>*,
.css-i3ef4m, .css-1knepg6, .Btn, .css-bda9ez, .pp-cc-name {
color: #5eded2;
text-shadow: 0 0 2px #5eded2;
font-family: 'Pirata One', cursive;
letter-spacing: 2px;
}
.css-o5an2m, .css-bda9ez, #search-input {
font-family: 'Pirata One', cursive;
letter-spacing: 2px;
font-size: 50px;
}
.css-1ciz3n, p, .css-1m151rf, .css-wjj1lc {
font-family: 'Crimson Text', serif;
letter-spacing: 0.5px;
}
/* Profile Ava customization*/
.css-1jj3srb:hover, .css-1jj3srb[data-hover], .css-bn1yyx:hover, .css-1bn1yyx[data-hover] {
color: #5eded2;}
.css-18bnokj {
width: 320px;
height: 450px;
clip-path: polygon(15% 0%, 85% 0%, 100% 15%, 100% 85%, 85% 100%, 15% 100%, 0% 85%, 0% 15%);
border: 2px groove #5eded2;
box-shadow: 0 0 15px #5eded2, inset 0 0 8px #5eded2;
position: relative;
}
/* Center Align */
.css-1uodvt1{
display: flex;
flex-flow: column;
}
/* Bot Card Holder*/
.css-1s5evre {
background: #080603;
border-radius: 10px;
position: relative;
z-index: 2;
transform: rotate(2deg);
transition: transform 0.3s ease;
gap: 0;
display: grid;
grid-template-columns: 1fr 1fr;
grid-template-rows: 50px 200px 1fr;
}
.css-1s5evre a{
display: contents;
}
.css-1s5evre:hover {
border-color: #5eded2;
color: #5eded2;
box-shadow:
0 0 8px #5eded2,
0 0 20px #5eded2,
inset 0 0 20px #5eded2;
text-shadow: 0 0 5px #5eded2;
transform: rotate(0deg);
}
.css-1gg9wg4, .css-1sxhvxh {
background: transparent;
}
.css-1wpvjoq, .css-1xdsfqv, .css-15vqpxh, .css-0, .pp-cc-wrapper,
.pp-cc-wrapper>*:not(.css-1s5evre) {
background: transparent !important;
}
.pp-cc-list-container{
display: flex;
flex-flow: row wrap;
justify-content: center;
}
.pp-cc-wrapper {
flex: 0 1 400px;
width: 400px;
height: 400px;
transition: all 300ms ease;
}
/* Bot Ava */
.css-1q7rmf0 ,.css-1q7rmf0>img, .css-1q7rmf0>video {
height: 250px;
border-radius: 5px;
grid-area: 2 / 1 / 2 / 1;
margin: 2px;
border-radius: 0;
overflow: visible;
}
/* Bot Ava Grey Scale Effect */
.css-1q7rmf0 {
-webkit-mask-image: linear-gradient(to bottom, #080603, transparent) !important;
mask-image: linear-gradient(to bottom, #080603 60%, transparent) !important;
-webkit-filter: grayscale(75%);
-webkit-transition: all .2s ease;
animation: shimmer 7s ease-in-out infinite;
}
.css-1q7rmf0:hover {
-webkit-filter: grayscale(0%);
transition: .7s ease;
animation: none;
}
/* The scanline overlay effect for Bot Ava */
.profile-character-card-avatar-aspect-ratio::after {
content: '';
position: absolute;
inset: 0;
background: linear-gradient(rgba(18, 16, 16, 0) 50%, rgba(0, 0, 0, 0.25) 50%), linear-gradient(90deg, rgba(255, 0, 0, 0.06), rgba(0, 255, 0, 0.02), rgba(0, 0, 255, 0.06));
background-size: 100% 5px, 4px 100%;
pointer-events: none;
z-index: 5;
opacity: 0.8;
transition: opacity 0.4s ease;
animation: scanlines-glitch 20s linear infinite;
}
/* Bot Name */
.pp-cc-name {
animation: flicker 5s ease-in-out infinite;
font-size: 20px;
padding: 1.5px;
width: 200%;
max-height: 90px;
height: 100%;
overflow: hidden auto;
white-space: normal;
word-wrap: break-word;
overflow-wrap: break-word;
text-align: center;
}
/* Bot description and tag appearance */
.css-199gcrh, .pp-cc-tags{
overflow: hidden auto;
}
.css-199gcrh {
grid-area: 2 / 2 / 2 / 2;
margin: 0.7rem;
}
.css-96l1id {
font-size: 0.9rem;
}
.pp-cc-tags{
grid-area: 3 / 1 / 3 / 3;
margin-top: 3rem;
animation: flicker 5s infinite;
}
.pp-cc-tags-item {
font-weight: 600;
padding: 5px 0 !important;
font-family: 'Crimson Text', serif;
display: inline-block !important;
min-height: 0 !important;
line-height: 1;
border: none !important;
border-bottom: 1px solid #2d2d2d ! important;
background: none ! important;
color: #FFF;
}
.css-k1urot {
gap: 6px;
justify-content: center;
align-items: flex-end;
}
/* Custom Tags content */
.css-199gcrh::after {
content: ' ';
text-decoration: underline;
text-decoration-thickness: 0.5px;
text-underline-offset: 5px;
position: absolute;
bottom: 120px;
right: 3px;
font-size: 1rem;
font-weight: 600;
line-height: 1;
}
.pp-cc-wrapper:has(.pp-tag-ordealofecstacy) .css-199gcrh::after {
content: 'Ordeal of Ecstacy';
font-family: 'Orbitron', 'Arial Black', sans-serif;
color: #ad2336;
text-shadow: none;
font-weight: 700;
}
.pp-cc-wrapper:has(.pp-tag-shrinebone) .css-199gcrh::after {
content: 'Shrinebone';
font-family: 'Beth Ellen', cursive;
color: #6d878a;
}
.pp-cc-wrapper:has(.pp-tag-hearthofbloomlock) .css-199gcrh::after {
content: 'The Hearth';
font-family: 'Monotype Corsiva', cursive;
letter-spacing: 3px;
color: #DC825B;
}
.pp-cc-wrapper:has(.pp-tag-demithel) .css-199gcrh::after {
content: 'Demi Domination';
}
/* Button customization */
.Btn2-purple {
background: transparent !important;
border-color: #5eded2;
}
/* The Star */
.css-1s5evre > div:first-of-type + div {
display: none;
}
/* Creator name */
.css-i3ef4m, pp-cc-creator-name, .css-1m151rf {
visibility: hidden;
display: none;
}
/* Token count customization */
.css-1c9wmts {
width: 50%;
right: 0;
left: unset;
}
/* Chat count customization */
._content_1chl9_11 {
background: transparent;
overflow: hidden;
}
.css-10cv7r2 {
position: absolute;
top: 35px;
}
.css-1ket5wn {
position: static;
filter: none;
width: 100%;
padding: 0;
}
.pp-cc-chats {
font-family: 'Share Tech Mono', monospace;
font-size: 0.8rem;
letter-spacing: 1px;
}
</style>
|