IAMCCS Cine FLF Engine Simple

FLF - first/last-frame - keyframing is how you keep an LTX shot on the rails: reference images locked at specific frames, motion interpolated between them. The IAMCCS Cine pack's fancier FLF nodes hide this behind boards and plans; Cine FLF Engine Simple is the compact version with the honest docstring: "Use timeline_data from Shotboard or compact manual_keyframes, not hundreds of visible widgets." It's the no-frills engine you use when you want the guide mechanism without the full Cine Shotboard ceremony.

It takes a batch of reference images and a short keyframe table, and injects those references into your conditioning and latent as guides at the frames you specify.

How it works

Feed it multi_input - a batched IMAGE of reference frames from a Cine Shotboard, a Reference Board, or anywhere you assemble them in order - and num_images to say how many slots to actually read. Then describe the schedule. You have two sources of truth:

• timeline_data - a synced Cine Shotboard FLF timeline; when connected it overrides manual keyframes so stage 2 follows the board.
• manual_keyframes - the compact fallback, and it's the format worth learning:

0 | 1 | 1.00
96 | 2 | 0.92
-1 | 5 | 0.88

Each line is frame | ref | strength (or second | ref | strength if insert_mode is seconds). ref is which image in the batch (1-indexed), strength is how hard to hold that frame, and -1 as the frame means "the last frame of the shot" - which is how you lock the final frame without knowing the length in advance. The frame_rate input matters when insert_mode is seconds.

The engine encodes each reference with the vae, adds it to the latent at the scheduled frame with a noise-mask, and layers the guide info into the conditioning. Outputs are positive, negative, and the guide-carrying latent.

Installing it

Part of IAMCCS-nodes:

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

Restart ComfyUI, or search "IAMCCS" in ComfyUI Manager. Baseline: ComfyUI ≥ 0.3.0, Python ≥ 3.12, Torch ≥ 2.8. No downloads beyond your LTX model + VAE.

The honest take

The manual_keyframes format is the heart of this node and it's genuinely ergonomic once you've written one table - five lines replace a graph of keyframing widgets. Where people get burned: num_images and the actual batch size disagreeing (a ref pointing past what you loaded silently does nothing), and mixing frames vs seconds modes mid-table. And if the Cine Shotboard is part of your workflow, feed it timeline_data and stop editing the table by hand - the whole point of this engine is that the board can drive it. If your shots are motion-heavy and guides drift, the Dyno sibling adds the extra attention-locking pass.