Nodes/ComfyUI-H3-Cast/MiniMax H3 Cast to Video (Extend)
ComfyUI Node

MiniMax H3 Cast to Video (Extend)

A ComfyUI node in MiniMax H3/cast with 29 inputs and 4 outputs.

By kat3ri·Created 16 days ago·Updated 9 days ago· 2
MiniMax H3 Cast to Video (Extend)
  • clip
  • vae
  • context_latent
  • audio_vae
  • cast_1
  • cast_2
  • cast_3
  • scene_images
  • first_frame
  • last_frame
  • world_latent
  • positive
  • latent
  • final_prompt
  • report
prompt
length124
max_views_per_member3
auto_introtrue
scene_description
include_voicestrue
context_frames2
context_strength1.00
context_static_timefalse
pin_last_frametrue
world_frames2
world_strength1.00
world_static_timefalse
ref_image_sizematch
ref_spacing1.0
ref_strength1.00
ref_decay0.00
ref_ramp0.0
CategoryMiniMax H3/cast

Inputs (29)

NameTypeDefaultDescription
clipCLIP
vaeVAE
context_latentLATENTAV latent output from a prior MiniMax H3 generation to continue from
promptSTRINGRefer to characters by NAME -- the <Picture i>/<Audio j> intro lines are written for you (see final_prompt output)
lengthINT1245–3600
max_views_per_memberINT31–9Views taken per cast member, in saved order. 9 total image slots are shared by all members + scene images.
auto_introBOOLEANtrueWrite the '<Picture 1>, <Picture 2>: Name -- description' intro lines automatically
audio_vaeoptVAE
cast_1optH3_CAST_MEMBER
cast_2optH3_CAST_MEMBER
cast_3optH3_CAST_MEMBER
scene_imagesoptIMAGEExtra reference images of the location/scene (e.g. room renders); batch = one ref slot per frame
scene_descriptionoptSTRINGWhat's actually in scene_images -- used verbatim in the intro line instead of the generic 'the location where this scene takes place.' placeholder.
include_voicesoptBOOLEANtrue
context_framesoptINT21–64Trailing latent frames of context_latent carried over as context
context_strengthoptFLOAT1.000–1
context_static_timeoptBOOLEANfalse
pin_last_frameoptBOOLEANtrue
first_frameoptIMAGEHard-pin this call's frame 0 (e.g. the prior clip's real last output frame) instead of pin_last_frame's decode
last_frameoptIMAGEPin this continuation segment's own final frame to an exact image, e.g. to land precisely on a known next shot/keyframe. Not supported by the fork's native MiniMaxH3VideoExtend -- silently dropped there (see report); supported when running against ComfyUI-MiniMax-H3-Extend's backport.
world_latentoptLATENT
world_framesoptINT21–64
world_strengthoptFLOAT1.000–1
world_static_timeoptBOOLEANfalse
ref_image_sizeoptCOMBOmatch2 options: match, max
ref_spacingoptFLOAT1.00–50
ref_strengthoptFLOAT1.000–1
ref_decayoptFLOAT0.000–1
ref_rampoptFLOAT0.00–50

Outputs (4)

NameTypeDescription
positiveCONDITIONING
latentLATENT
final_promptSTRING
reportSTRING