ComfyUI Node
Upscale using model adv (CRT)
Advanced upscaling with tiling, smart memory management, and precise output control
Upscale using model adv (CRT)
- image
- IMAGE
- width
- height
◄upscale_model_name▾►
◄use_fixed_resolutionfalse►
◄output_multiplier1.00►
◄fixed_width1024►
◄fixed_height1024►
◄tile_count1►
◄precisionauto►
◄batch_size1►
◄offload_modeltrue►
◄disable_cachefalse►
CategoryCRT/Image
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — | |
| upscale_model_name | COMBO | Select upscale model from models/upscale_models folder | |
| use_fixed_resolution | BOOLEAN | false | Enable to use fixed width/height instead of a multiplier |
| output_multiplier | FLOAT | 1.000.25–8 | Final output size multiplier relative to input (1.0=same as input, 2.0=double input size) |
| fixed_width | INT | 102464–8192 | Target width for the upscaled image |
| fixed_height | INT | 102464–8192 | Target height for the upscaled image |
| tile_count | COMBO | 1 | Number of tiles per side (1=no tiling, 4=4x4=16 tiles, 8=8x8=64 tiles, etc.) |
| precision | COMBO | auto | Processing precision (auto=fp16 on CUDA, fp32 on CPU) |
| batch_size | INT | 11–16 | Number of images to process simultaneously |
| offload_model | BOOLEAN | true | Offload model to CPU after processing to save VRAM |
| disable_cache | BOOLEAN | false | Disable caching to always reprocess images (useful for testing or varying results) |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | — |
| width | INT | — |
| height | INT | — |