Nodes/ComfyUI-FL-MiniMaxH3/FL_MiniMaxH3MotionPrepare
ComfyUI Node

FL_MiniMaxH3MotionPrepare

A ComfyUI node in FL/MiniMax H3/Internal with 12 inputs and 1 output.

By filliptm·Created about a month ago·Updated about 14 hours ago· 17
FL_MiniMaxH3MotionPrepare
  • vae
  • positive
  • latent
  • audio_vae
  • baseline_images
  • baseline_audio
  • shot_data
  • FL_H3_MOTION_PREP
target_long_side0
motion_coveragebalanced
max_hold4
expand_to_endtrue
upscale_methodlanczos
CategoryFL/MiniMax H3/Internal

Inputs (12)

NameTypeDefaultDescription
vaeVAE
target_long_sideINT00–163840 keeps source size. Otherwise upscale the long edge; dimensions align to 32 pixels.
motion_coverageCOMBObalancedOff benchmarks spatial refinement without temporal stretching. Uniform uses max_hold everywhere.
max_holdINT42–8Used only by uniform coverage. Adaptive coverage presets own their hold limits.
expand_to_endBOOLEANtrueExtend a short unheld tail through the last motion burst, matching H3 Time Smear.
upscale_methodCOMBOlanczos3 options: lanczos, bicubic, bilinear
positiveoptCONDITIONINGRequired without shot_plan. Shot-plan mode uses each shot's own conditioning instead.
latentoptLATENTCompleted native H3 video/audio latent, not an empty latent or temporal reshot.
audio_vaeoptVAERequired when the timeline expands; not needed with coverage off.
baseline_imagesoptIMAGEOptional shared decode of this exact source latent. Connect the same images used by the before panel.
baseline_audiooptAUDIOOptional shared decode of this exact source latent's soundtrack.
shot_dataoptFL_H3_MOTION_SHOT

Outputs (1)

NameTypeDescription
FL_H3_MOTION_PREPFL_H3_MOTION_PREP