ComfyUI Node
🧰 Ray's Model Tools: Queued Model Loader
Add diffusion models with + and remove rows with the trash button. One Run outputs a native MODEL list; standard downstream nodes process each model with the same singleton prompts, seeds and settings. Images appear together in one execution's preview. Rebatch Images combines the image list into a tensor batch. Each sampled model and its clones are offloaded through ComfyUI, including on error/cancel. DynamicVRAM also releases its RAM weight buffers; legacy loading can retain CPU weights in the execution cache. Precision is automatic. Loader options contains optional Spectrum autodetection. Keep compatible text encoders and VAE connected separately.
🧰 Ray's Model Tools: Queued Model Loader
- MODEL
◄model_name►
◄queued_models[]►
◄spectrum_enabledfalse►
Category👑 Ray/🧰 Model Tools
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| model_name | STRING | Select a registered diffusion model from the searchable folder dropdown. Search includes every subfolder. Precision and supported quantization are handled automatically by ComfyUI. | |
| queued_models | STRING | [] | Additional models in row order. One execution passes a native MODEL list through the workflow, reusing singleton prompts, seeds and settings. |
| spectrum_enabledopt | BOOLEAN | false | Enable optional Spectrum forecasting. Auto selects the installed adapter for each model. This approximation can change image/video results; unsupported models keep native sampling. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| MODEL | MODEL | Native ComfyUI MODEL list in selected row order. Downstream nodes process each model in one execution, reusing singleton prompts, seeds, and settings. |