IAMCCS Cine FLF Productor

The Cine FLF Engine Simple takes a hand-written keyframe table. The Cine FLF Productor is the production-grade version that takes an explicit guide plan - the JSON contract emitted by the Cine Filmmaker or CineInfo V2 breakout nodes - and turns it into the encoded references, conditioning, and latent for your LTX sampler. The source calls it "the production FLF backend driven by the explicit guide plan," and that's the right mental model: plan in, guided shot out.

You use this when the Shotboard has done the planning and you want the full plan respected, rather than retyping keyframes into the Simple engine.

How it works

It reads guide_plan_json (a JSON object with a guides array: each guide has a frame/second, a reference index, a strength, and optional motion force). If you leave guide_plan_json empty, it falls back to parsing timeline_data the same way the Shotboard does, so the node works even without the breakout. The reference images come through multi_input (batched, in order), encoded with the vae, and injected into the latent and conditioning at their scheduled frames. Two controls tune the result:

• strength_scale (0–2, default 1) - global multiplier over every guide's strength; the fast lever when a whole plan feels too strong or too weak.
• tail_safety_frames - reserve frames at the end so the final guide isn't sitting on the last sample.

Optional guide_data input lets a typed guide object feed in directly, and duration_seconds/frame_rate give the fallback timing when there's no plan.

Outputs

positive, negative, latent (all guide-applied) and a report describing what was injected - guides placed, strengths, reference count. Wire the three into your sampler.

Installing it

Part of IAMCCS-nodes:

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

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

The honest take

This node and the Cine Filmmaker Guide overlap - the Guide is a subclass of the Productor in the source - and you can't go wrong with either on the Cine-native FLF path; pick by whether your graph hands you a guide_plan_json (Productor) or a guide_data object (Guide). The genuinely useful pattern here is the fallback: because an empty guide_plan_json falls back to timeline_data, you can prototype with the Simple engine's table format and later feed a full plan without rewiring. The usual guide-strength advice applies - start strength_scale at 1.0 and dial toward 0.6–0.8 when motion dies - and tail_safety_frames is cheap insurance on long shots. Read the report; it's the fastest way to catch a reference index pointing at an image you didn't actually load.