wan21_interior_design

This is an interior design specialized model fine-tuned on the Wan2.1 base model, focusing on generating professional and spatially logical interior renderings.

The core trigger phrase is professional interior design photography with balanced lighting and architectural details—we recommend placing this in the first half of your prompt.

Recommended parameters:

Sampling Steps: 20-50 steps (20 steps is highly recommended for efficiency);

CFG Scale: 4-6 (avoids overfitting, maintains spatial coherence);

Weight: 0.8-1.0 (reduce to 0.6 if spatial distortions or proportion issues occur).

Recommended resolutions for this model: 960 and 1536.

For optimal results, use the res_2s + bong_tangent sampling combination; we also recommend using the WanVideo Tea Cache (native) node to significantly improve runtime speed.

The training dataset for this model was captioned using JoyCaption Beta One in Straightforward mode. We suggest you use the same method—upload images to JoyCaption Beta One and select Straightforward mode—to get accurate prompts for better results.