Choosing a model

Browse image and video model families and understand what changes when you switch.

The model list comes from the server, not from this documentation — what you see depends on your plan and your organization. Free accounts get a curated selection; VIP and SVIP accounts get all models.

Filtering the list

The video model selector, filtered by input type

Image models

Filter buttons group models into families: All, Nano Banana, GPT Image, Seedream, Unfiltered, and Other.

Video models

Two filter rows. The first is by input type — all, first frame, first and last frame, reference image, reference video, reference audio. The second is by family — Kling, Veo, Sora, Seedance, Wan, Hailuo, unfiltered, other.

The input-type filter is the useful one. If you need to animate a still you already have, filter to first frame and you only see models that can do it.

Each entry shows the model’s avatar, name, and a one-line description.

What a model’s capabilities look like

WJI does not list capabilities as badges. Instead, the controls change when you switch models. A setting that is not there is a setting that model does not support.

Switching models can also change your existing settings:

If the new model…	WJI…
Does not support your aspect ratio	Moves you to the closest ratio it does support
Accepts fewer reference images	Drops the extra ones from the front of the list
Does not accept references at all	Tells you the model does not support reference images
Has different duration limits	Applies its own default duration

The reference trimming is worth remembering. If you have five references and switch to a model that takes three, you keep the last three — the ones you added first are removed.

Switching models also clears the final prompt and re-optimizes it.

Aspect ratio fallbacks

Inside a workspace, if the new model does not support your current ratio, WJI first tries the workspace’s default ratio, and only then falls back to the model’s first supported ratio.

Model-specific settings

Anything unique to a model appears in the settings popover as a dynamic form. Common examples:

• Seed — -1 means unspecified
• Resolution or output size
• Chaos, quality, stylize, raw mode, HD on Midjourney-family models
• Task mode, auto duration, output format on Seedance video models
• Speaker, speech rate, pitch, loudness on text-to-speech models

There is no negative prompt field in WJI. Describe what you want rather than what you do not.