Nodes/ComfyUI-MiniMaxH3-Easy/MiniMax H3 Easy Media Splitter
ComfyUI Node

MiniMax H3 Easy Media Splitter

Back to plain IMAGE, VIDEO and AUDIO

By nkxx188·Created 28 days ago·Updated a day ago· 593
MiniMax H3 Easy Media Splitter
  • media_bundle
  • image_1
  • image_2
  • image_3
  • image_4
  • image_5
  • image_6
  • image_7
  • image_8
  • image_9
  • image_10
  • image_11
  • image_12
  • image_13
  • image_14
  • image_15
  • image_16
  • image_17
  • image_18
  • image_19
  • image_20
  • image_21
  • image_22
  • image_23
  • image_24
  • image_25
  • image_26
  • image_27
  • video_1
  • video_2
  • video_3
  • video_4
  • video_5
  • video_6
  • video_7
  • video_8
  • video_9
  • audio_1
  • audio_2
  • audio_3
  • audio_4
  • audio_5
  • audio_6
  • audio_7
  • audio_8
  • audio_9
image_count1
video_count0
audio_count0

Every MiniMax H3 Easy workflow funnels your images, videos, and audio through one custom MINIMAX_H3_MEDIA_BUNDLE type - the stuff the Media Loader and Media Bridge hand out. That's convenient inside the pack and useless outside it: a bundle is one opaque blob a normal ComfyUI node can't eat. MiniMax H3 Easy Media Splitter is the reverse utility that cracks it open again, emitting plain IMAGE, VIDEO, and AUDIO ports you can wire into anything.

Reach for it when you manage one shared media library in the Easy Media Loader but want some of that media to keep flowing into other parts of your graph - a reference image into an img2img chain, a clip into a different pack's video node, a voice sample into audio work. It's the same idea as the author's "four ways to provide media": Loader and Bridge collect into a bundle, Splitter expands one.

How it works

The node does not re-decode anything. It takes the bundle you hand it, groups its items by type, and re-packages them into the standard ComfyUI shapes: images come out as tensors, videos get wrapped into a normal VIDEO payload, audio into a {waveform, sample_rate} mapping. If a video item already carries decoded components, the splitter reuses them rather than decoding a second time.

The three count widgets are the interesting bit. They don't select which items come out - they decide how many output ports exist. Set image_count to 3 and you get image_1 through image_3; the rest of the 27 image ports stay hidden and don't clutter the canvas. The counts also don't have to match what's actually in the bundle. If a port would be empty, the node emits a silent execution blocker on it, so any branch wired to that spare port is skipped instead of erroring, and its populated siblings still run. Handy when you want to build a workflow with spare capacity before all the media is loaded.

The inputs and outputs that matter

  • media_bundle - a Media Bundle from MiniMax H3 Easy Media Loader or Media Bridge. Nothing else satisfies it; the node raises "Connect a Media Bundle from Media Loader or Media Bridge" otherwise.
  • image_count (default 1, max 27), video_count (default 0, max 9), audio_count (default 0, max 9) - set these to expose ports.
  • Outputs are image_1image_27, video_1video_9, audio_1audio_9, named in that order.

Installing it

MiniMax H3 Easy is one pack:

cd ComfyUI/custom_nodes
git clone https://github.com/nkxx188/ComfyUI-MiniMaxH3-Easy.git

Restart ComfyUI. In ComfyUI Manager you can search ComfyUI-MiniMaxH3-Easy instead - but pick the Nightly version when it asks, because the published releases lag the repo and Nightly is what stays current. You also need a recent ComfyUI build with the official MiniMax H3 nodes, since the whole pack imports comfy_extras.nodes_minimax_h3 under the hood. These five plumbing nodes load no models themselves, so to produce a bundle to split you still need the Easy Loader wired up with the H3 weights sitting in ComfyUI/models/diffusion_models, text_encoders, and vae.

Common issues

The usual beginner trip is feeding the splitter an image or a raw video instead of a bundle - it only accepts the Loader/Bridge output. One honest limitation to know: the splitter gives you no decoded-video cache. Of the four media methods in this pack, only the Media Loader caches decoded frames so re-runs skip re-decoding. If a split video gets used repeatedly, expect the decode cost each time - a preparation-time tax, not sampling time. For one-off handoffs to other workflows, none of that matters.

CategoryMiniMax H3 Easy

Inputs (4)

NameTypeDefaultDescription
media_bundleMINIMAX_H3_MEDIA_BUNDLE
image_countINT10–27
video_countINT00–9
audio_countINT00–9

Outputs (45)

NameTypeDescription
image_1*
image_2*
image_3*
image_4*
image_5*
image_6*
image_7*
image_8*
image_9*
image_10*
image_11*
image_12*
image_13*
image_14*
image_15*
image_16*
image_17*
image_18*
image_19*
image_20*
image_21*
image_22*
image_23*
image_24*
image_25*
image_26*
image_27*
video_1*
video_2*
video_3*
video_4*
video_5*
video_6*
video_7*
video_8*
video_9*
audio_1*
audio_2*
audio_3*
audio_4*
audio_5*
audio_6*
audio_7*
audio_8*
audio_9*