IAMCCS Cine LTX Sequencer Exact

IAMCCS Cine LTX Sequencer Exact is the pack's "baseline" node, and it says so right in its own description: "Exact IAMCCS cine port of the proven LTX FLF sequencer. Use this only as the behavior baseline before adding filmmaker UI." That's the whole personality of this node - it's the reference implementation. If you want to understand what the fancier Cine timeline nodes are actually doing under the hood, or you need predictable, boring keyframe insertion without a custom UI in the way, this is it.

It works on the same guide-insertion mechanism as its sibling (IAMCCS Cine LTX Sequencer): encode reference images with the video VAE and splice them into the latent at chosen frame positions, patching conditioning and noise mask. The difference is surface area. Where the FLF version parses a JSON/pipe timeline, this one gives you up to 50 explicit slots - insert_frame_1 … insert_frame_50, each with a matching strength_1 … strength_50, plus an optional insert_second_n twin for when you think in seconds instead of frames.

How it works (and why "Exact" matters)

This is a modern-format node - built on comfy_io.Schema (the newer ComfyUI node API) and it directly extends LTXVAddGuide from comfy_extras.nodes_lt, the native LTX guide node. So unlike most of this pack, it's not wrapping something into a black box; it is the native mechanism, given a fixed schema. That's the point of "exact": you get the proven behavior, unchanged, so you can validate that your workflow works before layering the prettier Cine tools on top.

Key controls:

• num_images - how many of the 50 slots are active. It's also the images_loaded display count.
• insert_mode - frames (default) or seconds. In seconds mode, insert_second_n gets converted using frame_rate.
• insert_frame_n - frame position for image n. -1 means the last frame, which is the classic FLF "end-frame conditioning" trick.
• strength_n - how hard that guide anchors the video (0–1).

It skips slots where the index exceeds the batch of multi_input images, and it asserts that no guide overruns the latent length rather than silently corrupting it. Outputs: patched positive, negative, and latent (with the tooltip straight from the author: "Video latent with added FLF guides").

Install

One node in IAMCCS-nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/IAMCCS/IAMCCS-nodes.git

or Manager → search IAMCCS. Because it builds on comfy_io and comfy_extras.nodes_lt, you need a recent ComfyUI - the README's floor (ComfyUI ≥ 0.3.0, Python ≥ 3.12, Torch ≥ 2.8) is not optional here; on an older build this node will fail to import.

Where people get burned

• Setting num_images past your loaded images. Slots with no image are silently skipped, which can look like "the node ignored my settings." Check the batch count on multi_input.
• Seconds vs. frames confusion. The two widget sets coexist; whichever insert_mode you pick, the other pair is ignored. If frames change but nothing moves, you're probably in seconds mode with second values of 0.
• The overrun assert. If a guide position is too close to the end of the latent, it errors loudly instead of trimming. That's deliberate. Reduce the frame position or lengthen the latent.
• "Exact" is a feature, not a bug. No timeline UI, no auto-parsing. If you're editing a scene with 20 keyframes, the fancier Cine nodes will be far less tedious - this node is for reproducibility and debugging.
• Zero search impressions, one-author tooling. But honestly, as a thin port of the native LTXVAddGuide, this is one of the more trustworthy nodes in the pack.