ComfyUI Node
Iris Sampler (Advanced)
A ComfyUI node in Iris-3B with 16 inputs and 1 output.
Iris Sampler (Advanced)
- model
- conditioning
- image
- IMAGE
◄steps50►
◄cfg3.0►
◄seed0►
◄batch_size1►
◄modetext to image►
◄denoise0.50►
◄shift4.0►
◄solver_order2►
◄cfg_start0.00►
◄cfg_end1.00►
◄width1024►
◄height1024►
◄keep_model_on_gpufalse►
CategoryIris-3B
Inputs (16)
| Name | Type | Default | Description |
|---|---|---|---|
| model | IRIS_MODEL | — | |
| conditioning | IRIS_CONDITIONING | — | |
| steps | INT | 501–500 | the model card default is 100 |
| cfg | FLOAT | 3.01–20 | — |
| seed | INT | 00–18446744073709550000 | — |
| batch_size | INT | 11–8 | images per run |
| mode | COMBO | text to image | image to image starts from the connected image instead of pure noise |
| denoise | FLOAT | 0.500.01–1 | image to image only: 1 ignores the image content, lower values keep more of it |
| shift | FLOAT | 4.00.1–20 | flow shift (trained with 4.0) |
| solver_order | INT | 21–2 | DPM-Solver++ order |
| cfg_start | FLOAT | 0.000–1 | CFG applies only for noise levels t in (cfg_start, cfg_end); 1 = pure noise |
| cfg_end | FLOAT | 1.000–1 | — |
| width | INT | 1024256–2048 | in image to image the input is scaled and cropped to this size |
| height | INT | 1024256–2048 | — |
| keep_model_on_gpu | BOOLEAN | false | skip moving the 6 GB DiT back to RAM after sampling |
| imageopt | IMAGE | source for image to image (ignored in text to image) |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | — |