Documentation
Set up Orme Models and install files into ComfyUI.
Install the CLI
Download the archive for your platform and
extract it. Open a terminal in the extracted directory and run the
CLI with ./orme-models. Keep aria2c beside
the executable when you move it to a directory on your
PATH.
./orme-models --help
You can compare the archive’s hash with the published SHA-256 checksums.
Configuration
Set the directory where ComfyUI looks for models. Examples below
assume orme-models is on your PATH.
orme-models config set models-dir "$HOME/ComfyUI/models"
orme-models config show
COMFYUI_MODELS_DIR overrides the saved path. Otherwise,
Orme uses the saved directory or looks for a ComfyUI checkout in the
current directory and its parents. Use
orme-models config unset models-dir to clear the saved
directory.
Install models
Pass each model’s path relative to the ComfyUI models directory. For
example, vae/wan_2.1_vae.safetensors installs under
ComfyUI/models/vae/.
Browse the public model index
for install paths. Orme reads its
JSON catalog by default.
Set ORME_MODELS_INDEX_URL to use a different catalog.
orme-models install \
vae/wan_2.1_vae.safetensors \
text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors
Orme processes up to two models at once. If you press Ctrl+C, run the same command again to resume matching partial downloads. Already verified models are skipped. Orme preserves an existing destination file if its content differs.
Resolve sources
resolve prints one download URL per line. The catalog
provides Hugging Face and can include a matching ModelScope mirror.
Set HF_ENDPOINT to change the Hugging Face download
host. The indexed file and its checksum still apply.
orme-models resolve vae/wan_2.1_vae.safetensors
Automation
Add --json to get one final JSON object on stdout.
Install results appear in result.items, in input order.
orme-models --json install vae/wan_2.1_vae.safetensors
| Exit code | Meaning |
|---|---|
| 0 | Success |
| 1 | Failure or partial success |
| 2 | Invalid arguments |
| 130 | Interrupted |