Model Parameters & Dynamic Controls

Different AI models expose different controls. The Model Parameters panel dynamically adapts to show only the parameters relevant to your selected model, giving you precise control without clutter.

Dynamic Parameter UI

The parameter panel responds to your model choice. Each model defines its own set of configurable parameters via a backend schema. The UI renders the appropriate controls based on this schema:

• Sliders — for numeric ranges like steps, guidance scale, or strength
• Ratio selectors — for aspect ratio or dimension choices
• Dropdown menus — for categorical options like scheduler or style

This adaptive approach keeps the interface clean while offering deep control when needed. Only the parameters relevant to the selected model are shown.

Seed Control

The seed determines the starting noise pattern for generation:

• Fixed seed — Use the same seed to reproduce identical results across generations
• Random seed — Each generation starts from a different noise pattern

Fixed seeds are essential for iterating on a specific look without starting over.

CFG Scale

Classifier-Free Guidance scale controls how strongly the generation follows your prompt. The available range and default value depend on the selected model. Higher values force the output to match your prompt more exactly but can reduce quality.

Scheduler Selection

When available, the scheduler controls the denoising step algorithm. Different schedulers affect speed, quality, and convergence. The available scheduler options depend on the selected model.

Note

For image tool panels (panorama, change angle, grid split, etc.), the model selector is fixed to the tool's assigned model. The dynamic parameters shown are specific to that tool's backend model configuration.