⭐ Star LTX Image Cut
Trim the junk frames off your LTX video batch before it gets anywhere near a sampler
- images
- images
- first_image
- last_image
- selected_frame
Anyone who's run LTX-Video long enough knows the drill: the output starts with a couple of mush frames while the model settles in, ends with another couple as it fades out, and your "10 second" clip is really 9 seconds of usable footage plus a thumbnail's worth of garbage. Star LTX Image Cut is the StarNodes node that trims those frames off a frame batch - and conveniently also hands you the first, last and a specific frame while it's at it.
It's in ⭐StarNodes/Video and works on the IMAGE-type frame batches that video decoders and the pack's LTXV all-in-one nodes output.
How it works
Three integer inputs slice the batch:
cut_first_frames(default 0) - how many frames to drop off the front.cut_last_frames(default 10) - how many to drop off the back. The author's default of 10 reflects the real-world observation that LTX tails are the worst part, though you'll want to tune it to your own generation.export_frame_number(default 0) - which frame to pull out as a singleselected_frameoutput (0-indexed, so 0 is the first remaining frame).
The cuts are clamped so you can't slice past the batch (it caps at total_frames - 1), which means it degrades gracefully instead of erroring on a short video.
The four outputs
images- the trimmed batch, ready to feed a VAE decode or the next stage.first_imageandlast_image- the first and last remaining content frames. These are the killer feature for LTX work: grab the last frame and feed it back into an image-to-video node to extend a clip, or use it as a keyframe for the first/last-frame workflows the pack's LTXV nodes support.selected_frame- whatever single frameexport_frame_numberpoints at, for cases where you want one specific moment.
Why it exists
This is a quality-of-life node in the best sense. LTX's 8n+1 frame snapping (the pack's own LTXV nodes snap lengths to that pattern) means you're often dealing with batches whose edges are near-guaranteed junk, and the standard fix - cutting frames by hand elsewhere in the graph - requires nodes that mostly don't exist in core ComfyUI. Here, trimming and frame-extraction are one small node, no scripting.
Installing it
Standard StarNodes install - ComfyUI Manager, search Starnodes, install, restart:
cd ComfyUI/custom_nodes
git clone https://github.com/Starnodes2024/ComfyUI_StarNodes
cd ComfyUI_StarNodes
pip install -r requirements.txt
Then double-click the canvas and search star - it lives under ⭐StarNodes/Video.
Gotchas
The default cut_last_frames = 10 is a lot - on a 97-frame clip that's over 10% of your footage gone. Tune it to your own outputs before you build it into a permanent workflow. And note the cuts happen before selected_frame is pulled, so export_frame_number is relative to the trimmed batch, not the original. If you're extending clips, wire last_image straight back into the pack's LTXV all-in-one as the first_frame - that's the loop this node was clearly built for.
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | — | |
| cut_first_frames | INT | 00–9999 | — |
| cut_last_frames | INT | 100–9999 | — |
| export_frame_number | INT | 00–9999 | — |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| first_image | IMAGE | — |
| last_image | IMAGE | — |
| selected_frame | IMAGE | — |