Model Option
While the core input and prompt process is consistent, each model provides unique configurable parameters. These settings allow you to fine-tune the AI's behavior, directly influencing the output's quality, style, and characteristics to meet your workflow's exact needs.
But if you're new to using AI and not ready to dive into detailed configurations, don't worry—the default setup usually delivers solid results.
Text Models
Web Search
Toggle on/off to allow the model to access external web information for its response.
Reasoning Level
Refers to how well the AI analyze, infer, and make logical connections between pieces of information to solve complex tasks. It's a measure of the model’s cognitive depth, beyond just recalling facts or generating fluent text.
Here’s what it typically implies:
Low: performs simple pattern matching, surface-level answers, and basic completions (e.g., auto-completing sentences, factual Q&A).
Medium: follow instructions, make basic inferences, and handle multi-step prompts with some success.
High: demonstrates strong logical thinking, can follow complex chains of reasoning, solve abstract problems, understand context deeply, and apply knowledge creatively—often similar to human critical thinking.
Temperature
Control the randomness or creativity of the output; higher values lead to more varied results. The value typically ranges from 0 to 1, though some systems allow values above 1.
Seed
A specific value that can be used to generate reproducible outputs for the same prompt, useful for iteration and debugging
Image Models
Image Size / Ratio
Define the dimensions or aspect ratio of the generated image (e.g. 4:3, 1:1)
Output Format
Select the desired file format for the output image (e.g. PNG, JPEG)
Output Compression
Control the visual fidelity and file size of the generated image
Background / Moderation
Specific settings related to background generation or content moderation, often auto-managed
CFC Scale (Classifier-free guidance scale)
Influence how strictly the AI adheres to your prompt High CFG scale (e.g., 12-16) The generated image will closely follow the text prompt, resulting in a more literal and detailed depiction of the described scene.
Default CFG scale (e.g., 7.5) Often, this value provides a good balance between prompt adherence and creative exploration. Low CFG scale (e.g., 1-5) The AI has more freedom to be creative, potentially leading to more abstract or unexpected results that may deviate from the exact prompt.
Num Inference Steps
Control the number of steps the AI takes to generate the image, impacting detail and generation time
Seed
A unique value to recreate specific image outputs with the same prompt
Negative Prompt
Specify elements you do not want to appear in the generated image
Video Models
Duration
Set the length of the generated video clip (e.g. 5 seconds)
Aspect Ratio
Define the video's screen proportions (e.g. 9:16, 16:9)
Resolution
Select the output video quality (e.g. 720p)
Style
For some models, choose a specific visual style for the video (e.g. cinematic, cartoon)
Seed
A unique value to help reproduce video outputs for specific prompts
Enhance Prompt
A toggle to allow the model to interpret and enhance your prompt for better results
Negative Prompt
Specify elements you do not want to appear in the generated video
CFC Scale (Classifier-free guidance scale)
Controls how much the video adheres to the prompt versus being creative.
By understanding and adjusting these parameters, you can precisely control the AI's output and tailor it to your specific workflow requirements.
Last updated