ComfyUI Node

BD Flame Fit

Fit the FLAME head model to a LANDMARKS_BATCH and emit a FACE_FIT — drop-in alternative to BD_FaceFit. FLAME is a proper morphable model: fitting its shape + expression basis produces a SUBJECT-ACCURATE face (the basis only spans real face shapes), unlike warping a neutral mesh. Full head — face + scalp + neck. Per view: Adam fits shape/expression/pose + a weak- perspective camera to the 105 MediaPipe landmarks via the official mediapipe_landmark_embedding. Requires the converted model (tools/convert_flame.py → flame2023_facewrap.npz) + mediapipe_landmark_embedding.npz.

By BizaNator·Created 7 months ago·Updated 16 days ago· 14
BD Flame Fit
  • landmarks_batch
  • face_fit
  • status
flame_npz_path
embedding_npz_path
shape_coeffs100
expr_coeffs50
iterations300
Category🧠BrainDead/FaceWrap

Inputs (6)

NameTypeDefaultDescription
landmarks_batchBD_LANDMARKS_BATCHOutput of BD_FaceLandmarks.
flame_npz_pathoptSTRINGPath to flame2023_facewrap.npz. Empty = /tmp/ComfyUI/models/flame/flame2023_facewrap.npz
embedding_npz_pathoptSTRINGPath to mediapipe_landmark_embedding.npz. Empty = /tmp/ComfyUI/models/flame/mediapipe_landmark_embedding.npz
shape_coeffsoptINT10010–300Number of FLAME shape basis components to fit. More = finer identity, slower, more overfit risk.
expr_coeffsoptINT500–100Number of FLAME expression components to fit.
iterationsoptINT30050–2000Adam iterations per view.

Outputs (2)

NameTypeDescription
face_fitBD_FACE_FIT
statusSTRING