Chatting in FictionLab
Last updated
Last updated
Chatting in FictionLab is very easy, but this guide will explain every feature to allow you to maximize your roleplay experience!
The chat is between you and the AI. The AI acts as the other characters while you act as your own character or a pre-defined character, depending on the scenario settings.
You can write your responses to the input field at the bottom or choose from two suggested responses if you're too lazy to write your own. The suggested choices can be disabled from the light bulb next to the send icon.
At the top navigation bar, there are three buttons. The first one from the left is the sharing button which allows you to send the scenario to your friends. You can even share a private scenario, allowing people with the link to access it. The middle button is for favoriting the scenario. This makes it appear in the "Favorites" section on your user page. Favoriting scenarios allows you to find them easily.
The last button is the settings button. Clicking it will open a side menu with a bunch of settings allowing you to customize the chat to your liking.
The LLM model selection allows you to choose between the AI models that FictionLab offers. Each model has a unique "personality" and writing style, so you can experiment with them and see what fits your roleplaying style the best. Below are all the current models:
Parameters: 70B, 32K context
Description: The default free AI model of FictionLab. Good creativity and decent memory. Completely unlimited and unfiltered!
Parameters: 72B, 128K context
Description: Glendora V3 is the new balanced model for FictionLab+. Amazing coherence and creativity, especially with detailed scenarios and multiple characters.
Parameters: 671B, 128K context
Notes: This model is experimental. It is based on DeepSeek V3 0324.
Description: Experimental and possibly limited-time model! Based on the latest DeepSeek V3. Blends razor-sharp coherence with a wildly unpredictable edge. It crafts stories with masterful precision—then flips the script when you least expect it.
Parameters: 70B, 32K context
Description: Aurelius is trained to focus more on small details in the roleplay. Great for slow burn and slice-of-life roleplays.
Note: Messages sent to this model might be used for training purposes.
Parameters: 235B, 41K context
Description: A master of mystery, intrigue, and hidden meanings. Riddleheart weaves clues and tension into every line, blending slow-burn suspense with emotional depth. Ideal for mysteries, conspiracies, and romance-driven roleplays.
Parameters: 253B, 131K context
Description: Overhauled version of Sorcerer V3. Based on Nemotron Ultra. Very intelligent model that takes instructions to heart. Great for scenarios where instruction following is important. Not as uncensored as other models.
Parameters: 70B, 32K context
Description: This model excels at running darker roleplays. Perfect for horror or violent roleplays where survival isn't always guaranteed.
Stream Text Response - Enables a typing animation where the generated text is shown to the user in real time. With this setting enabled, the response will only appear when the response is completed. Stream Text Response doesn't work for iOS devices at this time.
Show Response Suggestions - Enables or disables the two pre-generated choices after receiving a response from the bot.
The user character is the character you, the user, will play as during the scenario. You can access the user character menu from the side menu and edit it as the story goes on.
Hovering over the user character cards brings up three dots. Clicking on these dots allows you to either delete or edit the character card.
User character cards should be written in the same way as normal character cards are. Describe the looks and personality in the description. More detail usually gives better results.
Free users are limited to 10 user character slots. FictionLab+ users have unlimited character slots.
Custom LLM instructions are almost identical to custom instructions in the scenario creation page. You can use it to guide or instruct the AI to act how you want. Instructions should be clear and concise for the best results. The limit for Custom LLM Instructions for free users is 200 characters and 800 for FictionLab+ users.
Custom LLM instructions should be used to nudge the AI in a certain direction and deleted after the desired results are seen.
The narration voice setting allows you to change the voice that reads messages when using the narration feature. You can activate narration by clicking on the speaker icon while hovering above the AI's responses.
Chat background styles allow you to customize the chat window to fit your roleplay better and make your roleplay more immersive. Free users have access to 7 backgrounds while FictionLab+ users have access to over 40 different backgrounds.
FictionLab uses a variety of different symbols to indicate different behaviors. For example, actions are usually wrapped in *asterisks* while dialogue is wrapped in "quotes like these". FictionLab also supports bold letters in AI responses. Bold text is activated with **double asterisks**. It is usually used to emphasize actions or dialogue.
Hovering over a message makes three vertical dots appear on the side of the message. Clicking on these dots makes a message menu appear. This message menu allows you to edit, rewind, and delete messages. You can also regenerate the message using this menu.
Editing the message allows you to change parts of the message to your liking. It can be used to change events that have happened during the roleplay and to drive the roleplay into a different direction. Use the previously mentioned formatting for best results when editing.
Messages can also be reported if you find them to break our TOS and guidelines. Our staff team will then review the message and take action if needed.
Messages can be regenerated if you're not happy with the results. You can press the reload symbol that appears at the bottom of the message while hovering to regenerate the message. After regenerating, you can switch between the response versions using the arrow symbols at the bottom of the message box.
