Nodes/ComfyUI-H3-Multishot-Advance/Multishot Advance Controls
ComfyUI Node

Multishot Advance Controls

A ComfyUI node in video/minimax with 14 inputs and 9 outputs.

By KursatAs·Created 8 days ago·Updated about 19 hours ago· 1
Multishot Advance Controls
  • model
  • width
  • height
  • frames_per_shot
  • seed
  • steps
  • sampler_name
  • scheduler
  • shot_count
  • use_file_prompts
aspect_ratio3:4 (Portrait Standard)
megapixels1.0
multiple32
frames_per_shot243
seed0
steps8
sampler_nameer_sde
schedulerbeta
shot_count0
use_file_promptsfalse
take_seconds0.0
windowauto
pin_frames22
Categoryvideo/minimax

Inputs (14)

NameTypeDefaultDescription
aspect_ratioCOMBO3:4 (Portrait Standard)Canvas aspect ratio. Replaces the separate Resolution Selector node.
megapixelsFLOAT1.00.1–16Target total megapixels. 1.0 MP with 3:4 and multiple 32 gives 896x1184.
multipleINT3232–256Snap width and height to this multiple. 32 is the H3-safe default.
frames_per_shotINT2435–481H3's 17k+5 grid at 24fps. 243 = ~10.1s default; 362 = ~15.1s trained max. Drives stage A's length AND stage B's per-shot length.
seedINT00–18446744073709550000Seed sent to the H3 sampler. Default is 0 and the after-generate mode defaults to fixed.
stepsINT81–50
sampler_nameCOMBOer_sde44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38
schedulerCOMBObeta9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3
shot_countoptINT00–30How many shots the scene is. 0 = one shot per prompt in the script (or lets prompt planning decide). Wire this to BOTH the sampler's shot_count and the prompt generator's num_shots so they cannot disagree.
use_file_promptsoptBOOLEANfalseLegacy prompt-source flag. OFF = manual prompt entry; ON = external prompt-set source when the workflow provides its own switch.
take_secondsoptFLOAT0.00–600EXTEND TAKE: 0 = off. Any other value = the length of the finished take; frames_per_shot and shot_count above are then OVERRIDDEN by a window sized for this card (see 'window') and the count that fills the time. Set prompt planning to 'extend take' so it writes ONE speech across the windows.
windowoptCOMBOautoFrames per window when take_seconds is set. auto = the largest window whose estimated activation pool fits with most of the weights resident on THIS card (wire model for a real weight size; 15 GB assumed otherwise). Bigger window = fewer joins; smaller = less VRAM.
modeloptMODELOptional, EXTEND TAKE only: the loaded H3 model, so 'auto' can size its weights.
pin_framesoptINT220–56EXTEND TAKE only: set this to the same value as the memory sampler's pin_frames. 22 is the tested default. If you change the sampler pin but leave this at 22, take_seconds will calculate the wrong delivered length.

Outputs (9)

NameTypeDescription
widthINT
heightINT
frames_per_shotINT
seedINT
stepsINT
sampler_nameSTRING
schedulerSTRING
shot_countINT
use_file_promptsBOOLEAN