IAMCCS Cine Filmmaker Backend

The Cine Shotboard plans the shot; IAMCCS Cine Filmmaker Backend is where the plan turns into something an LTX sampler can actually run. You hand it a loaded model, a clip, and the cine_linx plan, and it hands back the patched model, positive/negative conditioning, the video and audio latents, guide data, and a combined audio track - everything the sampling chain needs, wired to match the plan. If the FLF Productor and Guide nodes are the "how the guides get applied" side, this is the "give me a ready pipeline" side.

It's the runtime counterpart to Cine Filmmaker: that node extracts the plan, this one consumes it. The pack calls it "the Filmmaker runtime backend that consumes the FLFreal visual timeline trunk."

How it works

It takes your base LTX model and clip, reads the plan out of cine_linx, and produces conditioned positive/negative, a video_latent, and an audio_latent (via the optional audio_vae - LTX-2's synchronized audio path needs a VAE that can handle audio). use_custom_audio toggles whether your own timeline audio gets encoded instead of the model's generated audio. The promptrelay_safety_budget (and its enable toggle) caps the conditioning relay so long prompt stacks don't blow past a sane size - default 24000. goya_audio_strict is the one with a real tooltip: "When enabled, custom timeline audio encode problems raise instead of silently falling back to empty audio." If you've ever lost audio silently, you know why that toggle exists.

Inputs and outputs that matter

Required: model, clip, cine_linx. Optional but commonly used: audio_vae (for audio output), use_custom_audio, and the safety budget.

Outputs: model (patched), positive/negative (conditioning), video_latent, audio_latent, guide_data, frame_rate (float), combined_audio, report. Wire the latents into the sampler, guide_data into the guide nodes, and combined_audio into VAE-decode/audio-out.

Installing it

Part of IAMCCS-nodes:

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

Restart ComfyUI (or Manager → search "IAMCCS"). Baseline: ComfyUI ≥ 0.3.0, Python ≥ 3.12, Torch ≥ 2.8. No model downloads from the pack itself - you supply the LTX model and VAE; the audio VAE is part of your LTX-2 setup, not something this node fetches.

Where people get burned

Two silent-failure traps. First, the audio path: if use_custom_audio is on but your audio_vae isn't connected (or your audio won't encode), the old behavior was to silently fall back to empty audio - that's precisely what goya_audio_strict exists to surface. Turn it on when you're debugging missing audio. Second, the pack's newest LTX-2.3/V2V backend instructions live behind the author's Patreon, so if this node's behavior doesn't match the README, check you're on the latest release. And remember the report output - it's a JSON run-down of what the backend actually did, and it's the first place to look when a "worked in theory" shot comes back empty.