CogvideoX Image to Video ComfyUI workflow

Use this ComfyUI workflow to turn a starter image and text describing motion into a CogVideoX video, with a simple upscaling option.

An included CogVideoX motion LoRA is intended to guide camera movement.

The workflow also includes a simple method for upscaling its video output.

The ZIP includes the workflow JSON, a starter image, and a creation PNG that also contains the workflow.

Expand the collapsed Note nodes for model links and instructions on where to place the models.

Sugestão

The listed ComfyUI packs are ComfyUI-CogVideoXWrapper, comfy_mtb, comfy-image-saver, ComfyUI-Custom-Scripts, ComfyUI-KJNodes, mikey_nodes, and comfyui-various.

Listed model requirements include microsoft/Florence-2-large, 4x-UltraSharp.pth, and t5xxl_fp8_e4m3fn.safetensors.

The workflow also lists alibaba-pai/CogVideoX-Fun-V1.1-5b-InP and pytorch_lora_weights.safetensors; expand the Note nodes for model links and placement details.

Add movement details to the prompt; the creator says CogVideoX relies on motion information in text.

Sugestão

The description gives Fun 1.1 as fixed at 16:10 and 720×480, while Fun 1.5 goes up to double size.

The creator reports that the included motion LoRA works with Fun 1.1 but not Fun 1.5.

The creator could not run it on an 8 GB card; they expect 12 GB may work but have not tested it, and currently render on a 16 GB card.