SCAIL-2

SCAIL-2

Repackaged SCAIL-2 video model files for ComfyUI, built on Wan 2.1 14B, with character replacement workflows. For ComfyUI users who swap characters in video.

Made by Comfy-Org on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: This repository holds SCAIL-2 model files repackaged by Comfy-Org so they load in ComfyUI. The original model comes from zai-org. The names show a Wan 2.1 14B video model in five precisions, plus two LoRA files, and the card links two character replacement workflows.

Who it suits: ComfyUI users who want to replace a character in a video clip on their own computer. The card is short and gives file locations and workflows, so it assumes you already know ComfyUI. Use only characters, faces and footage you have the right to use.

Title: Who uses it, and how
Industry | Who uses it | How they use it | Example
Video making | Hobby creators | Swap the character in a short clip | Replace a mascot in your own animation test
Animation | Indie animators | Try a new character design in existing motion | Reuse one motion clip with a different design
AI research | ComfyUI tinkerers | Compare the model file precisions | Run the fp8 and int8 workflows on one clip
Post production | Editors | Prototype a replacement before real work | Test a cartoon character on a rough cut

Skip it if you do not use ComfyUI, because the files are packaged for that program. Do not use it to put a real person into video without their permission.

Where to use it:

• ComfyUI, on Windows, macOS or Linux
• The two character replacement workflows the card links
• Windows, macOS or Linux for downloading the files, with Python

Use case | Who | What they get back
Character replacement | Creators | A clip where a chosen character replaces the original
LoRA testing | Tinkerers | Output changed by the DPO or relight LoRA
Precision tests | Researchers | Output from different model file versions

How to set it up: You need ComfyUI installed, plus Python and pip to download. Then you copy each file into the matching folder inside ComfyUI's models directory.

• ComfyUI
• Python and pip
• Disk space for the model files you choose

1. Install the Hugging Face hub library.
2. Download the repository.
3. Copy the files into ComfyUI folders, then open a workflow.

pip install -U huggingface_hub
hf download Comfy-Org/SCAIL-2

Title: Where each kind of file goes
Folder | Files
diffusion_models | wan2.1_14B_SCAIL_2 in fp16, fp8_scaled, int8_convrot, mxfp8 and an nvfp4 mix
loras | wan2.1_SCAIL_2_DPO_lora_bf16 and wan2.1_SCAIL_2_relight_lora_bf16

How to use it: Open one of the two character replacement workflows in ComfyUI, load your source clip and a reference character, and run it. Use the int8 workflow with the int8 model file, and pick a precision that fits your memory.

Tool | What it does
SCAIL-2 Character Replacement workflow | Replaces a character in a clip, shown before and after
SCAIL-2 Int8 workflow | Runs the same task with the int8 model file
DPO LoRA | An extra file the card ships, without saying what it changes
Relight LoRA | An extra file whose name points to lighting changes

Example: You open the SCAIL-2 character replacement workflow, load a short clip of your own and a reference character you own, and run it. The workflow returns a clip with the new character. The card shows before and after thumbnails.

For newcomers and experts: If you are new, learn the basics of ComfyUI first, then try the fp8 or int8 file on a clip of a few seconds. Check the result before you try longer ones.

For people in the field, the file names show a 14B Wan 2.1 base in fp16, fp8 scaled, int8 convrot and two 4-bit or 8-bit float mixes. The card does not explain the variants or give benchmark figures.

Download the repository
Copy files into ComfyUI folders
Open a character replacement workflow
Load a clip and a reference character
Get the edited clip

Good to know:

• License: MIT
• Cost to run: none named, you use your own computer
• Data it reaches: it downloads files from Hugging Face, and the card names nothing else it touches
• Not for: swapping in a real person without their consent, or use outside ComfyUI
• The card gives no benchmark scores or hardware notes

On Hugging Face, 13 November 2026

• Library: diffusion-single-file
• Downloads in the last 30 days: 696,480
• Likes: 141
• License: mit
• Based on: zai-org/SCAIL-2
• Tags: diffusion-single-file, comfyui

Links

• Model: https://huggingface.co/Comfy-Org/SCAIL-2
• Files: https://huggingface.co/Comfy-Org/SCAIL-2/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download Comfy-Org/SCAIL-2

Then follow the usage in its model card. Nothing in this kit runs by itself.