bytedance-research

/

Lance

like 445

Follow

bytedance-research 1.39k

Hi,

Thanks for your interest!

The main difference between the two models is their training focus and intended use cases.

Lance_3B_Video is the more general unified model. It supports the full set of capabilities, including video/image generation, understanding, and editing. It does have image-related capabilities, but its training is primarily focused on video tasks, so video generation/editing/understanding is the main strength of this model.

Lance_3B, on the other hand, is trained only with image-related data. As a result, it is more specialized for image tasks and generally performs better than the more general video model on image generation/editing/understanding.

So in short:

• Use Lance_3B_Video if you need video capabilities, while still having access to image-related functions.
• Use Lance_3B if your main focus is image tasks and you want stronger image performance.

Hope this clarifies the difference!