🔧Generation Settings

Generation Settings allow you to fine tune specific aspects of your SpicyChat experience, such as the creativity of character responses, which AI model is used to generate messages, and how many tokens are utilized in chats. You can adjust these settings at any time during a conversation.

To modify your settings:

1. Select the three dots at the top right corner of any conversation.

2. Choose the “Generation Settings” option.

3. Change the Inference Model or tweak the Inference Settings to match your preferences.

Some settings are only available to premium subscribers.

What Are Inference Models?

The Inference Model refers to which AI model the character is using to generate responses in the conversation. For full details on AI models and the ones available on SpicyChat, visit the AI Models page.

What Are Inference Settings?

Inference Settings are advanced options that allow you to experiment with different variables for generating character responses. Adjusting Inference Settings is available only to True Supporter and I’m All In subscribers.

What Is Temperature?

This setting controls how tame or wild a character's responses will be. By default, Temperature is set to 0.7 and can be adjusted from 0 to 1.5.

• At a low temperature, the character will play it safe, choosing more predictable answers that are reliable and clear. This is ideal for users who want the conversation to stay focused and consistent.

• At a high temperature, characters become more adventurous, encouraging more diverse and creative responses. This can lead to unexpected conversations that can be fun and surprising, but may be less precise and coherent.

What Is Top-P?

Top-P (or nucleus sampling) is a setting that controls how much a character explores its options before deciding which words or phrases to use when crafting their responses.

It’s a great way to balance between keeping a conversation on track and letting characters think outside the box, giving you the freedom to choose how creative or focused you want the conversation to be.

• When set to a lower value, the character focuses on a smaller set of words or phrases, resulting in responses that tend to be more stable and reliable.

• A higher value expands the character's creativity, giving them access to a larger pool of words and phrases to choose from. This provides more variety, creativity, and unexpected responses.

The default value for Top-P is 0.7, with options to set it between 0.01 and 1.

What Is Top-K?

Top-K allows users to limit how many word options characters can consider. It allows you to control the range of a character's vocabulary, helping you strike the right balance between precision and imagination.

• With a lower Top-K value, the character limits their vocabulary to fewer words, making responses more focused and reliable. This helps to reduce the chance of generating replies that don’t make sense.

• A higher Top-K value gives characters a broader selection of words to pick from, encouraging interesting and unexpected responses. This comes at the cost of characters potentially using sentences that are strange and difficult to understand.

The default value for Top-K is 90, with a range from 1 to 100.

What Does Response Max Tokens Do?

This setting lets you fine-tune how a character responds. By default, responses are capped at 180 tokens, but True Supporters and I’m All In subscribers can increase this to a maximum of 300 tokens for longer, more detailed replies.

Keep in mind that a higher token limit doesn’t guarantee longer or complete responses. The character may still give shorter answers depending on the conversation.

To learn more about tokens and how they function, view the full explanation here.