[Legacy] H3 Continuum Assemble V3.2.4

H3 Continuum Assemble V3.2.4 is the "before" picture. It's the assembler that shipped with the V3.2.4 release of the H3 Continuum pack, and the V3.3 update kept it registered only so old saved workflows keep loading. If your workflow references it, it still works - but if you're wiring up a new graph, use H3 Continuum Assemble + Seam instead, which is this node plus guarded video seam correction.

What it does is the same core job as its successor: take the chunk images and audio that you decoded with ComfyUI Core VAE Decode nodes, trim the decoded continuation overlap, align the audio on cumulative frame boundaries, and optionally apply Audio Seam Auto at decoded audio boundaries. assembly_plan from the sampler tells it exactly how the chunks fit together - which frames overlap, where the true chunk boundaries are, and the exact total duration.

Inputs and outputs

• images - decoded chunk images (list).
• audio - decoded chunk audio (list).
• assembly_plan - from any Continuum V3/V3.2.4/V3.3 sampler's assembly_plan output.
• exact_total_duration (true) - trims or minimally pads to the requested total.
• audio_seam (Auto) - corrects decoded audio boundaries only; video frames are never altered.
• diagnostics (Basic) - report verbosity only.

Outputs: images, audio (ready for Create Video / save), and report (STRING, readable via Core Preview as Text).

Notice what's missing compared to the current node: no video_seam control at all. The V3.2.4 assembler corrects audio seams but leaves decoded video boundaries alone - that's precisely the gap the V3.3 "Assemble + Seam" node fills with its guarded transient/micro-flash correction.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/ukr8b3g-cmyk/ComfyUI-H3-Continuum.git

or via ComfyUI Manager ("H3 Continuum"), then restart. No extra pip packages; you need a working MiniMax H3 environment and ComfyUI v0.32.0+.

If you still see a video_seam-less assembler in a workflow and are tempted to swap it for the new node: they're drop-in compatible on images/audio/assembly_plan, so the swap is trivial. The one thing not to do is leave this node in place and add the new one - each expects the full decoded chunk list, not a partial assembly. If your saved V3.2.4 workflow runs fine, though, there's no fire: it just means you're on the older seam behavior until you rewire.