The Graph
A trained LoRA carries the identity into the model and the text encoder. IP-Adapter FaceID Plus V2 takes a single reference photo and holds the face on top of it. An OpenPose ControlNet takes the pose from a second photo, with its own strength setting, so a weak pose hint can be turned down without touching the face. Variants cover SD 1.5 and SDXL, with and without the pose hint.
Training the LoRA
A Colab notebook runs kohya_ss sd-scripts with Accelerate in fp16. Captioned images come from Google Drive, and each epoch's LoRA is saved back to it. It starts at 512 pixels, network dim 16 and a cosine schedule, with notes for when free Colab runs out of memory.
A Note on the Image
The graph is drawn from the real workflow JSON: every node, setting and link. LoRA names, photo file names and the prompt were replaced, because they referred to real people.
The same pipeline later became part of a full product: see FaceSearch.
Need the same face in every image?
ComfyUI graphs, LoRA training and FaceID set up for your characters, models or brand. Baydot builds the workflow and hands it over documented.
Book a Free Call