IAMCCS Final Audio Mixer

Once your LTX video has a dialogue track and a foley track, you need them as one clean audio signal for the final encode. IAMCCS Final Audio Mixer is the two-input audio mixer that does the job: dialogue in one port, optional foley in the other, per-track gain in dB, peak normalization on the way out. No audio workstation, no dozen-node chain - it's the "finish the audio here" node at the end of the IAMCCS Cine audio path.

The name says "Final" and it means it - this is designed to be the last audio node before your VAE decode or video combine. It pairs naturally with the Dialogue Timing Reconciler (which measures your dialogue) and whatever generates your foley.

How it works

It reads both waveforms, applies your gain values in decibels, pads and channel-matches them to a common length, and sums them. dialogue_gain_db defaults to 0 (leave dialogue untouched); foley_gain_db defaults to −6, which is the sensible starting point since foley should sit under the dialogue, not over it. If normalize_peak is on (default), the mixed result is scaled down only when it would clip - peaks above 1.0 get divided back under. The code is careful about one thing in particular: if the sample rates of the two tracks don't match, it raises a clear error instead of silently producing garbage - a genuinely helpful failure mode.

Inputs and outputs

• dialogue_audio (AUDIO, required) and foley_audio (AUDIO, optional - leave it disconnected for dialogue-only).
• dialogue_gain_db, foley_gain_db, normalize_peak - the three knobs.

Outputs: final_audio (AUDIO, ready for encode) and a report (JSON) with sample rate, sample counts, and peak-before-normalize - useful when you're trying to figure out why the mix sounds quiet.

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. Audio handling imports torchaudio lazily, so if ComfyUI already does audio in your install you're fine; no model files.

Where people get burned

The sample-rate mismatch error is the one to know: if your dialogue TTS and your foley generator ran at different sample rates, this node won't silently resample - it raises. Fix it at the source (resample one track upstream) rather than expecting the mixer to paper over it. Also remember that normalize_peak only acts when the peak is over 1.0 - a quiet mix stays quiet, because this is a safety limiter, not an auto-loudness tool. If your final audio is noticeably quieter than the dialogue alone, that's the −6 dB foley default doing its job, not a bug. One light criticism: it's a two-track mixer, nothing more - if you need music beds or multi-layer mixes, this isn't the node, and the pack's bigger AudioBoard is the heavier tool for that.