IAMCCS Cine Filmmaker Guide
Push reference frames into your shot with one strength knob
- positive
- negative
- vae
- latent
- guide_data
- positive
- negative
- latent
- report
Reference-guided video generation is how you hold a shot together: you lock certain frames to keyframe images and the model animates between them. IAMCCS Cine Filmmaker Guide is the node that applies a plan's guide data to your LTX conditioning and latent - the place where a shotboard's reference images stop being a plan and start steering the sampling. In source it's a subclass of the Cine FLF Productor, so it inherits the same guide-application machinery but presents it as a single-generation applicator with just two knobs.
It's part of the IAMCCS Cine stack, and it's the node you place after the Cine Filmmaker (or CineInfo) breakout and before the sampler.
How it works
You give it the positive/negative conditioning, a vae (used to encode the guide images into latent space), the video latent, and the guide_data object that the breakout nodes emit. It encodes the reference images, applies them as guide latents into the latent and conditioning at their scheduled frames, then hands everything back. Two controls tune it:
strength_scale(0–2, default 1) - global multiplier on guide strength. Drop it toward 0.5 when the references dominate and the motion freezes; push it up when the shot drifts away from your frames.tail_safety_frames- extra frames reserved at the end so the last guide doesn't sit on the very last sample. Set it a few frames above zero on long shots where the tail gets clipped.
Outputs
positive, negative, latent (all with the guides applied), plus a report describing what got injected. Wire the outputs straight into the 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 and VAE.
The honest take
Guide strength is the eternal knob fight in this style of generation: too weak and your references are ignored, too strong and the video becomes a slideshow between locked frames. strength_scale is the single most useful control here because it rescales an entire plan at once - start at 1.0, and if motion is dead, cut it to 0.6–0.8 rather than editing every keyframe. The tail_safety_frames knob exists because a guide scheduled at the last frame tends to get eaten; a small value is cheap insurance. If you're running the classic LTX Director-compatible path, the Guide 1:1 variant is the closer match; this one is the Cine-native FLF applicator. Either way, the report output will tell you exactly which frames got what strength - read it before blaming the sampler.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| positive | CONDITIONING | — | |
| negative | CONDITIONING | — | |
| vae | VAE | — | |
| latent | LATENT | — | |
| guide_data | GUIDE_DATA | — | |
| strength_scale | FLOAT | 1.000–2 | — |
| tail_safety_frames | INT | 00–240 | — |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| positive | CONDITIONING | — |
| negative | CONDITIONING | — |
| latent | LATENT | — |
| report | STRING | — |