-
-
Notifications
You must be signed in to change notification settings - Fork 448
Detailer
Detailer is an extra workflow task that can be enabled for any text/image/control generation by selecting it from the menu
What does detailer do?
Using specified model(s) and following the specified settings, detect objects in the image Settings that apply are:
- Minimum confidence of detected object
- Maximum number of detected objects
- Max overlap between detected objects
- Minimum and Maximum size of detected objects
- Crops detected object from the image and resizes it to resolution ideal for loaded model
- Runs inpaint operation for each cropped object using specified rules
- Resizes inpainted object to original size and places it back in the original image
Settings that apply are:
- Prompt and Negative prompt
- Steps and Strength of inpaint operation
- Edge padding and Edge blur
If detailer prompt is specified, it will be used, otherwise primary prompt will be used
If you're using multiple detailer models, you can specify prompt for each model separately by placing them on separate lines: 1st model will use prompt from 1st line, etc.
Most object detection models are trained on a single class of objects, e.g. face or hands
Some models are trained on multiple classes in which case you can specify which classes to detect
Example multi-class model is yolo11m model
When using multi-class model, available classes will be displayed in the log when you run the model
SD.Next comes with several pre-defined detailer single-class models:
face-yolo8n, hand_yolov8n, person_yolov8n-seg, eyes-v1, eyes-full-v1
And one multi-class model:
yolo11m
Additional models can be added by downloading pretrained model and placing it into folder
specified in Settings -> System paths -> Yolo models (default: models/yolo
Compatible models are any from Yolo model family
Models from different families such as Mediapipe are not supported