H3 Continuous - Trim Rendered Output (Legacy)

H3 Continuous - Trim Rendered Output is the simplest node in this pack: it cuts a rendered H3 clip's images (and audio, if you feed it) from the head, the tail, or both. Set head_trim_frames and tail_trim_frames, get a shorter clip back. That's it. No boundary analysis, no handover metadata, no latent math.

The reason it exists is that early versions of this H3 continuation work needed manual trimming, and it hung around for backwards compatibility and quick eyeball-edits. You can safely ignore it in the shipped v1.4 workflows - they use the smarter Output / Stitch nodes that do boundary-aware trimming. Where this node is genuinely useful: when you just rendered a clip, want to shave a few frames off the front or back for preview, and don't care about seam mechanics.

The part that will bite you

It says it right in the node's own description: this node never changes the saved AV latent or the continuation node's latent handover window. It's a rendered-output-only trim. That's a feature when you're previewing, and a trap when you're not.

Imagine you trim a clip's tail by 10 frames manually, then run the next continuation using that clip's latent. The continuation still happens at the full latent boundary - your visible trim and the latent handover now disagree, and your seam will show it. tail_trim_frames here is also deliberately independent of landing_tail_frames in the Continue node. The v1.2+ suite nodes exist precisely to keep rendered output and latent handover in lockstep; this one is the escape hatch that assumes you know what you're doing.

What it does under the hood

Two small behaviors worth knowing:

• 0/0 is a true bypass. If both trims are zero, the node returns the exact input untouched - including the audio. (In a very early version it still length-normalized the audio at 0/0; that's been fixed, so it's genuinely transparent now.)
• Audio is duration-matched, not just sliced. It trims the audio to exactly match the kept video frames. This cleans up H3's small per-clip audio-grid rounding drift, so the audio doesn't run a few milliseconds long after a trim.

It will raise an error if head_trim_frames + tail_trim_frames >= the clip's frame count (it won't happily delete the whole clip), and it's hardcoded around H3's 24 fps native rate via the fps input - leave that at 24 unless you know your render is different.

The one-sentence summary

It's a clean, honest trimmer that belongs in the "old stuff that still works" drawer. Use it for previews and quick cuts. If you're stitching a multi-clip H3 chain and want your trims to line up with the continuation boundary, reach for the Output / Stitch nodes from the same pack instead, and keep this one unplugged.

Installation is the whole pack: ComfyUI Manager (search Herrgotts-H3-Infinite-Continuation-Suite) or

cd ComfyUI/custom_nodes
git clone https://github.com/HerrgottMargott/Herrgotts-H3-Infinite-Continuation-Suite.git

then restart. No extra dependencies, no model downloads beyond the standard MiniMax H3 setup.