ComfyUI Node
h4 - Identity Engine
A ComfyUI node in h4/IdentityEngine with 38 inputs and 4 outputs.
h4 - Identity Engine
- face_image
- model_opt
- clip_opt
- vae_opt
- image_optional
- character_image
- model
- clip
- vae
◄presetNone►
◄ckpt_name▾►
◄vae_name▾►
◄clip_name▾►
◄seed0►
◄steps20►
◄cfg8.00►
◄sampler_name▾►
◄scheduler▾►
◄denoise1.00►
◄positive_dna►
◄positive—►
◄negative—►
◄width512►
◄height512►
◄batch_size1►
◄face_model▾►
◄swap_enabledtrue►
◄face_similarity0.50►
◄restore_enabledtrue►
◄restore_visibility1.00►
◄restore_modelcodeformer-v0.1.0.pth►
◄codeformer_weight0.50►
◄swap_modelinswapper_128.onnx►
◄boost_enabledfalse►
◄occlusion_enabledtrue►
◄sam_modelnone►
◄preserve_glassestrue►
◄upscale_enabledfalse►
◄upscale_modelnone►
◄upscale_face_onlyfalse►
◄positive_text—►
◄negative_text—►
Categoryh4/IdentityEngine
Inputs (38)
| Name | Type | Default | Description |
|---|---|---|---|
| preset | STRING | None | Select a saved character preset. |
| ckpt_name | COMBO | The primary Checkpoint. Determines the art style. | |
| vae_name | COMBO | Color correction. 'Baked' means use what's inside the Checkpoint. | |
| clip_name | COMBO | Text Interpreter. 'Use Checkpoint' is usually best. | |
| seed | INT | 00–18446744073709550000 | The magic number definition. Same Seed + Same Settings = Same Image. |
| steps | INT | 201–10000 | How many times to polish the image. 20-30 is standard. |
| cfg | FLOAT | 8.000–100 | Creativity Scale. Low (5-7) = Creative. High (8-12) = Strict. |
| sampler_name | COMBO | The drawing method. 'euler' is fast, 'dpmpp_2m' is detailed. | |
| scheduler | COMBO | How noise is removed. 'normal' or 'karras' work well. | |
| denoise | FLOAT | 1.000–1 | Img2Img Only: How much to change the original? 1.0 = New Image, 0.4 = Minor Edits. |
| positive_dna | STRING | DNA Prompt: Fixed character persistence (e.g. 'blonde, scars'). Concats with Positive. | |
| positive | STRING | Scene Prompt: What they are doing (e.g. 'riding a moto'). | |
| negative | STRING | What do you NOT want? (e.g. 'blurry, ugly, bad hands') | |
| width | INT | 51216–4096 | Image Width. |
| height | INT | 51216–4096 | Image Height. |
| batch_size | INT | 11–64 | How many images to create at once. |
| face_image | IMAGE | Drag a photo here to clone this face onto the character. | |
| face_model | COMBO | Or pick a pre-saved .safetensors face model. 'None' to disable. | |
| swap_enabled | BOOLEAN | true | Turn ON to enable face swapping. |
| face_similarity | FLOAT | 0.500–1 | How strictly to copy the face features. Higher = More likeness, but maybe weird. |
| restore_enabled | BOOLEAN | true | Fix blurry faces automatically. |
| restore_visibility | FLOAT | 1.000–1 | Restore strength. 1.0 = Full Fix. |
| restore_model | COMBO | codeformer-v0.1.0.pth | The repair tool to use. |
| codeformer_weight | FLOAT | 0.500–1 | Fidelity vs. Quality. 0.0 = High Fidelity, 1.0 = Smooth Skin. |
| swap_model | COMBO | inswapper_128.onnx | The logic engine for the swap. |
| boost_enabled | BOOLEAN | false | High-Res swap? Slower but sharper. |
| occlusion_enabled | BOOLEAN | true | Prevent face from overlapping hair/hands. |
| sam_model | COMBO | none | SAM model for perfect occlusion masks. |
| preserve_glasses | BOOLEAN | true | Attempt to keep the target's eyewear. |
| upscale_enabled | BOOLEAN | false | Upscale the FINAL result. |
| upscale_model | COMBO | none | The upscaler engine. |
| upscale_face_only | BOOLEAN | false | Only upscale the face region? (Saves VRAM) |
| model_optopt | MODEL | — | |
| clip_optopt | CLIP | — | |
| vae_optopt | VAE | — | |
| image_optionalopt | IMAGE | — | |
| positive_textopt | STRING | [Override] Connect external Scene Prompt here (e.g. from a Primitive). | |
| negative_textopt | STRING | [Override] Connect external Negative Prompt here. |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| character_image | IMAGE | — |
| model | MODEL | — |
| clip | CLIP | — |
| vae | VAE | — |