Use the LoRA Library in Cinema Motion

The LoRA Library in Cinema Motion is unavailable. It ran on WAN 2.2, and that engine has been retired from Workroom. We're bringing video character consistency back on newer models — we're not announcing a date yet. The LoRA Library in Picture Lab is unaffected.

The LoRA Library in Cinema Motion stored your trained avatars so you could apply them directly to AI video generation. WAN 2.2 was the only video model with a dedicated LoRA Library — the character stayed consistent in the output without a reference image from Picture Lab. Here's how it worked.

Select WAN 2.2

You opened Cinema Motion and chose WAN 2.2 from the model selector. This was the model with a dedicated LoRA Library control in the Cinema Motion bar.

Open the LoRA Library

Clicking LoRA Library opened a panel on the left side with three tabs: Avatar, Style, and Product.

LoRA Library modal with Avatar, Style, and Product tabs open in Cinema Motion alongside the WAN 2.2 interface

Avatar — trained characters (people or consistent personas)
Style — trained visual styles
Product — trained product appearances

If you hadn't trained anything yet, the tabs showed "No scenes found". See Train an avatar LoRA for the training workflow.

Apply a LoRA

You clicked a trained avatar in the Avatar tab. Selecting an avatar closed the panel and activated the LoRA for the next generation. You wrote your prompt describing the scene, then clicked Generate, and the character from your LoRA appeared in the generated video.

The prompt worked best when it referenced the character naturally — describing the scene and action, not the person's appearance. The LoRA handled consistency on its own.

Why video LoRAs were different

In Picture Lab, LoRAs generate a reference image that Cinema Motion then animates. That two-step process introduces variation at the animation stage. WAN 2.2 with a LoRA skipped the intermediate image entirely — the character identity went straight into the video model, giving more consistent results across multiple generations.