LTX Sequencer (Automated)

AnotherLTXSequencer is the "stop telling me how many images I have" version of LTX guide injection. Where the pack's LTXVMultiGuide asks you to manage positions and counts by hand, this node just asks for a batch of images and an index list - it reads how many images you actually supplied from the list itself and injects each one at the frame position the list says. Same family of trick as its sibling, but pointed squarely at automation.

Like LTXVMultiGuide, this is a mirror of the community-famous WhatDreamsCost LTX Sequencer, and it uses only the reliable path: loop over the reference images, VAE-encode each, convert the target pixel frame to a latent index using the VAE's downscale formula, and append it as a keyframe via LTXVAddGuide.append_keyframe. No manual attention entries, clean latent dict on the way out. The whole point of that design is staying stable across ComfyUI updates, and it's the right call for a node that lives in a video model ecosystem that changes fast.

The inputs are the LTX stack plus the automation controls. positive, negative, vae, latent pass through and come back updated. multi_input is your batch of reference images, indices is the list of frame positions (one per image), num_images (0–50, default 1) says how many to process - but here's the automation kicker: if your indices list has more than one entry, the node overrides num_images and uses the length of the list instead, capped at the batch size. You can practically stop thinking about the count. insert_mode (frames or seconds) picks the unit for your indices, frame_rate (default 24) converts seconds to frames, and strength (0–1, default 1) is the global guide weight applied to every injected frame.

Where you'd reach for it: any workflow where the number of keyframes changes between runs - a batch of storyboard frames that varies in length, a generated sequence whose keyframe count you don't know in advance, or a node-driven graph where a human shouldn't have to edit a count widget. The "automated" in its name is earned: wire in a variable-length batch and an index list and it just works, which is genuinely nicer than most LTX keyframing nodes that demand you keep counts and positions in sync by hand.

Gotchas: the same LTX-core requirement as its sibling - it imports from comfy_extras.nodes_lt, so a current ComfyUI is the real dependency, not the pack's own install. And a couple of things to double-check if output looks wrong: positions are treated as frame numbers by default (switch insert_mode to seconds if you think in time), and if your index list is shorter than your batch, trailing images are silently skipped. The node also prints chatty debug lines to the console on every run - harmless, ignorable.

Install: ComfyUI Manager → search "AnotherUtils", or cd ComfyUI/custom_nodes && git clone https://github.com/marcoc2/ComfyUI-AnotherUtils, restart. No extra Python packages, no model downloads - it uses the LTX nodes and VAE you already have.