Model
Trends
.ai
Model
Trends
.ai
the best open source ai models

FaceDetailer (Impact Pack) node in ComfyUI

FaceDetailer finds faces (or hands, eyes) with a YOLO detector, crops each one, re-renders it at full resolution with your model and prompt, and pastes it back, fixing the small, mushy faces in wide shots.
Node pack: ComfyUI-Impact-Pack
Category: Popular custom node packs

What FaceDetailer (Impact Pack) does

FaceDetailer, from Dr. Lt. Data's ComfyUI-Impact-Pack, automates the manual "inpaint the face at higher resolution" routine. An Ultralytics detector (face_yolov8m.pt, hand_yolov8s.pt) finds regions, SAM optionally refines the mask, each region is cropped with guide_size, upscaled, sampled with denoise 0.4-0.5 and composited back with feathering. The node bundles the whole pipeline: detector, sampler settings and compositing in one place.
It takes the same model, clip, vae and conditioning as your main sampler, plus a BBOX_DETECTOR from UltralyticsDetectorProvider and optionally a SAM_MODEL. Outputs include the fixed image, the cropped-refined previews and the masks, which is handy for checking what was detected. For several passes (face, then hands) chain two FaceDetailer nodes.

Inputs

Name
Type
What it is
image
IMAGE
The rendered picture.
model / clip / vae / positive / negative
MODEL, CLIP, VAE, CONDITIONING
Same as the main sampler.
bbox_detector
BBOX_DETECTOR
From UltralyticsDetectorProvider with face_yolov8m.pt or similar.
sam_model_opt
SAM_MODEL
Optional SAM for tighter masks.
guide_size / max_size
INT
Crop is upscaled to guide_size (384-512 for SD1.5, 1024 for SDXL) and capped at max_size.
seed / steps / cfg / sampler / scheduler / denoise
sampler settings
denoise 0.35-0.5 is typical.
feather / noise_mask / force_inpaint
INT / BOOLEAN
Edge blending and inpaint behaviour.
bbox_threshold / bbox_dilation / bbox_crop_factor
FLOAT / INT
Detection confidence, mask growth and context around the face (crop_factor 3 recommended).

Outputs

Name
Type
What it is
image
IMAGE
The picture with refined faces.
cropped_refined
IMAGE
Each refined crop, for inspection.
cropped_enhanced_alpha
IMAGE
Crops with alpha.
mask
MASK
The combined detection mask.
detailer_pipe
DETAILER_PIPE
Bundled settings for chained detailers.
AD
Skip the setup: ComfyUI workflows preinstalled on the cloud
BitVector runs ready-made ComfyUI workflows on its own GPUs. No install, no missing nodes, no red boxes. Open it on your phone or laptop and generate in a minute.

How to use FaceDetailer (Impact Pack)

  1. Install ComfyUI-Impact-Pack (and Impact Subpack for the Ultralytics provider) from ComfyUI Manager.
  2. Add UltralyticsDetectorProvider and pick face_yolov8m.pt (downloaded to models/ultralytics/bbox).
  3. Add FaceDetailer; connect image from VAE Decode, model/clip/vae/positive/negative from the main graph, bbox_detector from the provider.
  4. Set guide_size 512 (SD1.5) or 1024 (SDXL), denoise 0.45, crop_factor 3, feather 5.
  5. Connect the image output to Save Image; check cropped_refined with a Preview Image.

Settings and tips

  • Use a prompt focused on the face ("detailed face, sharp eyes") through a separate CLIP Text Encode when the main prompt is about the scene.
  • For many small faces raise bbox_threshold to skip the background crowd, or lower it when a face is missed.
  • Hands: hand_yolov8s.pt with denoise 0.5 and a "perfect hands" prompt; results vary.
  • Chain FaceDetailer -> FaceDetailer with the detailer_pipe output to run face then eyes.
  • A LoRA for the character in the face pass keeps identity consistent.

Troubleshooting FaceDetailer (Impact Pack)

No face detected, image passes through unchanged

Why it happens
bbox_threshold too high, the face is tiny, or the detector model failed to download.

How to fix it
Lower bbox_threshold to 0.3, make sure face_yolov8m.pt exists in models/ultralytics/bbox, and upscale the picture before the detailer.

The new face does not match the body (colour or style)

Why it happens
denoise too high, or a different prompt/LoRA in the face pass.

How to fix it
denoise 0.3-0.45, crop_factor 3 for context, and the same model and LoRAs as the main pass.

Visible square around the face

Why it happens
feather too low or noise_mask off with a hard mask.

How to fix it
feather 10-20, noise_mask on, bbox_dilation 10.

ModuleNotFoundError: ultralytics / Impact Subpack missing

Why it happens
UltralyticsDetectorProvider lives in the Impact Subpack, which installs separately on new versions.

How to fix it
Install ComfyUI-Impact-Subpack from Manager and restart.

Faces get a second set of eyes or a different person

Why it happens
Denoise 0.6+ redraws the face from the prompt rather than refining it.

How to fix it
Lower denoise; use a FaceID IP-Adapter or a character LoRA for identity.

AD
Run this workflow from your phone
Every workflow on this page is preinstalled on BitVector with the models and custom nodes already in place. Pick one, type a prompt, done. Zero setup, nothing to download.

Questions about FaceDetailer (Impact Pack)

Is FaceDetailer the same as ADetailer?

It is the ComfyUI equivalent of the A1111 ADetailer extension, with the same detector models.

Does it work with Flux?

Yes, with cfg 1.0 and Flux Guidance in the positive. Use guide_size 1024.

Can it fix hands?

It can refine hands with the hand detector; results depend on the model. MeshGraphormer hand refiner in ControlNet Aux is an alternative.

Related nodes

KSampler
ComfyUI core
KSampler runs the denoising loop: it takes the model, prompts and a latent and produces the finished latent image, controlled by seed, steps, cfg, sampler, scheduler and denoise.
Upscale Image (using Model)
ComfyUI core
Upscale Image (using Model) runs an ESRGAN-style upscaler (4x-UltraSharp, RealESRGAN, 4x_NMKD-Siax, 4x_foolhardy_Remacri) over an IMAGE and returns it enlarged by the model's fixed factor with added detail.
IPAdapter (IP-Adapter Plus)
ComfyUI_IPAdapter_plus
The IPAdapter nodes feed a reference picture into the model as an "image prompt", transferring style, composition or a face without describing them in words; IPAdapter Unified Loader plus IPAdapter Advanced are the usual pair.
Grow Mask, Feather Mask and Invert Mask
ComfyUI core
The core mask tools expand (Grow Mask), soften (Feather Mask) and flip (Invert Mask) a MASK so inpainted regions blend into the surrounding picture without visible seams.
ReActor Fast Face Swap
ComfyUI-ReActor
ReActor swaps a face from a source photo onto the generated picture with InsightFace, with optional restoration (GFPGAN, CodeFormer) and face boosting, for consistent characters without training a LoRA.
AD
Tired of fixing nodes? Let the cloud do it
BitVector keeps hundreds of ComfyUI workflows installed, updated and tested on fast cloud GPUs. No Python, no CUDA errors, no VRAM limits. Works from any browser.

More ComfyUI nodes

CLIP Text Encode (Prompt)
ComfyUI core
CLIP Text Encode turns a text prompt into CONDITIONING using the model text encoder. One node holds the positive prompt, a second one the negative prompt.
ComfyUI Manager
ComfyUI-Manager
ComfyUI Manager is the extension that installs, updates and fixes custom node packs and models from inside the interface, resolves missing nodes in imported workflows and snapshots your setup.
Load Checkpoint
ComfyUI core
Load Checkpoint (CheckpointLoaderSimple) opens a .safetensors or .ckpt model file and hands out the three parts every workflow needs: the diffusion MODEL, the CLIP text encoder and the VAE.
Save Image
ComfyUI core
Save Image writes the IMAGE tensor to ComfyUI/output as a PNG, with the whole workflow embedded in the file metadata so the picture can be dragged back into ComfyUI to restore the graph.
VAE Decode
ComfyUI core
VAE Decode converts the sampled LATENT into a pixel IMAGE with the VAE; VAE Decode (Tiled) does the same in tiles for very large images.
Empty Latent Image
ComfyUI core
Empty Latent Image creates the blank latent canvas (width, height, batch_size) that text-to-image sampling starts from; dimensions must be multiples of 8 and match the model family.