ComfyUI Node
h4 - Face Detailer
A ComfyUI node in h4/IdentityEngine with 17 inputs and 2 outputs.
h4 - Face Detailer
- image
- model
- clip
- vae
- bbox_detector
- detailed_image
- face_crop
◄guide_size512►
◄guide_size_fortrue►
◄max_megapixels2.0►
◄steps20►
◄cfg8.00►
◄denoise0.30►
◄sampler_name▾►
◄scheduler▾►
◄positivedetailed face, high qualty, skin texture, pore details►
◄negativeblur, smooth, cartoon, anime, lowres►
◄feather_mask32►
◄force_inpainttrue►
Categoryh4/IdentityEngine
Inputs (17)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | The upscaled image to refine. | |
| model | MODEL | The Model/LoRA stack used to generate the details. | |
| clip | CLIP | CLIP model for text guidance. | |
| vae | VAE | VAE for encoding/decoding the face crop. | |
| guide_size | INT | 512256–2048 | Resolution to process the face at. Higher = More detail but slower. 512-768 is sweet spot. |
| guide_size_for | BOOLEAN | true | Reference size for the bounding box or the actual crop? usually BBox is safer. |
| max_megapixels | FLOAT | 2.00.5–16 | Safety limit for VRAM usage. |
| steps | INT | 201–100 | Sampler steps for the detail pass. |
| cfg | FLOAT | 8.000–100 | Creativity scale. |
| denoise | FLOAT | 0.300.01–1 | How much to hallucinate? 0.30 adds texture without changing structure. Above 0.5 will change the face. |
| sampler_name | COMBO | 44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38 | |
| scheduler | COMBO | 9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3 | |
| positive | STRING | detailed face, high qualty, skin texture, pore details | Prompts to guide the detailing. Keep it simple. |
| negative | STRING | blur, smooth, cartoon, anime, lowres | What to avoid. |
| feather_mask | INT | 320–256 | Softness of the paste edge in pixels. |
| force_inpaint | BOOLEAN | true | Use inpaint masking logic (better blending) vs simple overlay. |
| bbox_detectoropt | BBOX_DETECTOR | Optional external detector (not used yet, internal detector is default). |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| detailed_image | IMAGE | — |
| face_crop | IMAGE | — |