comfy-videogen
comfy-videogen
Use this skill for video generation through the comfy-videogen CLI. Local LTX
2.3 modes use model files under /mnt/models/comfyui. Remote Seedance 2.0 modes
use ComfyUI API Nodes vendored by comfy-diffusion and require
COMFY_ORG_API_KEY.
If a built-in LTX 2.3 profile is missing files, use comfy-model-downloader for
the requested local videogen.<mode> capability before running inference. Do
not use the downloader for Seedance 2.0; seedance2-api has no local model
files.
The CLI is quiet by default and prints only final JSON. Use --verbose only when
debugging ComfyUI runtime output, warnings, or progress bars.
If comfy-videogen or comfy-models is not available, use
comfy-tools-setup first. In this repository, prefer uv run comfy-videogen;
outside the repo, let comfy-tools-setup install the CLIs with uv tool.