Search Results
6/9/2025, 7:52:38 PM
>>42258575
>multiple Scenarios with the same character card for different stories?
branch them
>Say I want a sfw chat and a separate nsfw chat for the same character.
branch them. then open the branches in MANAGE CHAT FILES option
or create a NEW CHAT for this card
>What's the correct way to have multiple First Messages
you do a NEW CHAT for this card, rewrite first message there and start conversation way you want. then when you click a NEW CHAT again - you will get a chat with original first message. edit it again whatever you want. your previous chat is saved in MANAGE CHAT FILES option
>when I want to play with different prompt settings and character descriptions?
different character description - duplicate card then
different prompt settings - either author's note and attach it to this chat, or if this is complicated - create a new prompt in preset and toggle depending on chat
>What does ooc stand for? Are there any other important commands I should know?
OOC - out of character. it is a term from IRL roleplaying. it is NOT a special command. AI were trained on examples of roleplay (chats and logs), and real people give each other commands and details via OOC. then another person follows them. because of this AI learned that OOC in context of roleplay stands for "it is a direct instruction of what you must do next". it is artifact of AI training that works in our favor. you can write (please do: xxx) or [TASK - XXX] or **CommmANDDD 4 AI --- * , they all the same with various degree of effectiveness. OOC is just what anons do the most because it is the most convenient
>I noticed that gpt will do decent quality nsfw fairly easily
it is really a problem of model or context or preset. hard to say without seeing what exactly you are doing and typing. and again GPT is like the worst model to start AI-hobby with. too gimmick for newfrens. Deepseek, Gemini, Claude - anything else if more easy to master and get a hold of
>What does using a jailbroken llm through SillyTavern provide that a regular ChatGPT session can't?
full control over preset. in your chatGPT session there is a system prompt set by developers that you cannot disable and extra moderation system. they are absent on API used in ST. in addition a model control. chatGPT auto switches you between various versions of GPT depending on your prompt. in ST you always use ONLY model you specifically selected
>multiple Scenarios with the same character card for different stories?
branch them
>Say I want a sfw chat and a separate nsfw chat for the same character.
branch them. then open the branches in MANAGE CHAT FILES option
or create a NEW CHAT for this card
>What's the correct way to have multiple First Messages
you do a NEW CHAT for this card, rewrite first message there and start conversation way you want. then when you click a NEW CHAT again - you will get a chat with original first message. edit it again whatever you want. your previous chat is saved in MANAGE CHAT FILES option
>when I want to play with different prompt settings and character descriptions?
different character description - duplicate card then
different prompt settings - either author's note and attach it to this chat, or if this is complicated - create a new prompt in preset and toggle depending on chat
>What does ooc stand for? Are there any other important commands I should know?
OOC - out of character. it is a term from IRL roleplaying. it is NOT a special command. AI were trained on examples of roleplay (chats and logs), and real people give each other commands and details via OOC. then another person follows them. because of this AI learned that OOC in context of roleplay stands for "it is a direct instruction of what you must do next". it is artifact of AI training that works in our favor. you can write (please do: xxx) or [TASK - XXX] or **CommmANDDD 4 AI --- * , they all the same with various degree of effectiveness. OOC is just what anons do the most because it is the most convenient
>I noticed that gpt will do decent quality nsfw fairly easily
it is really a problem of model or context or preset. hard to say without seeing what exactly you are doing and typing. and again GPT is like the worst model to start AI-hobby with. too gimmick for newfrens. Deepseek, Gemini, Claude - anything else if more easy to master and get a hold of
>What does using a jailbroken llm through SillyTavern provide that a regular ChatGPT session can't?
full control over preset. in your chatGPT session there is a system prompt set by developers that you cannot disable and extra moderation system. they are absent on API used in ST. in addition a model control. chatGPT auto switches you between various versions of GPT depending on your prompt. in ST you always use ONLY model you specifically selected
Page 1