Question 1

How does VIDU Reference-to-Image Q2 maintain consistency with reference images?

Accepted Answer

The model uses advanced deep learning techniques to capture and preserve key characteristics such as structure and style from the reference inputs, ensuring that the generated image reflects the core identity of the original.

Question 2

What types of prompts work best with this model?

Accepted Answer

Detailed and descriptive text prompts yield the best results. Including specific directions regarding style, composition, and desired variations helps the model align the final output with your vision.

Question 3

Are there any limitations on the number of reference images?

Accepted Answer

Yes, you can provide up to 7 reference images. This allows you to combine multiple sources of inspiration while ensuring the model can process them effectively.

Question 4

What output resolutions are available?

Accepted Answer

You can choose from 1k, 2k, or 4k resolutions. The selection depends on your project's quality requirements and output medium.