FlexlatentAudioBlend (DEPRECATED USE FLEX LATENT BLEND) ⚡🅡🅞🅣🅘

Let's lead with the part that actually matters: the node's own display name says "(DEPRECATED USE FLEX LATENT BLEND)". If you're starting a new workflow, don't reach for this one - use RyanOnTheInside's Flex Latent Blend node instead. This page exists because the node still runs, still shows up in older saved workflows, and you deserve to know what it does and why it's on its way out.

Worth noting it also sits in a different category path than the rest of this pack's current nodes - latent/audio rather than the RyanOnTheInside/FlexFeatures/... tree everything actively maintained lives under. That mismatch is itself a signal: this predates the pack's current Feature/Flex architecture, and it never got moved into it.

What it did

It's a latent blender with an audio-flavored name but nothing audio-specific actually wired into its inputs - no AUDIO type anywhere in its schema. It takes two latents, latent1 and latent2, and mixes them by blend_factors (0–1, default 0.5) using one of five blend_mode options - normal (straight linear crossfade), add, subtract, multiply, or overlay - with interpolation set to either linear or step for how the blend factor is applied. Output is a single blended LATENT.

This is exactly the kind of thing RyanOnTheInside's own Flex system was built to make reactive - a blend factor that shifts frame by frame based on an extracted feature, rather than sitting at one fixed ratio for the whole sequence - but this particular node predates that pattern. It has no feature_param, feature_mode, feature_threshold, or strength inputs, no optional feature socket at all. It's a static blend, full stop, which is almost certainly the actual reason it got superseded: everything else in the pack got rebuilt around the reactive Feature pipeline, and this one node was left behind using the old, non-reactive design.

Why Flex Latent Blend replaces it

The successor node the deprecation notice points to follows the pack's current pattern - the same strength / feature_threshold / feature_param / feature_mode / opt_feature set you'll find on every FlexImage* and FlexAudio* node in this pack. That means a latent blend that can be driven by audio amplitude, motion, or any other extracted feature - the thing this node's name implies but was never actually built to do.

How to install it

If you're here because an existing workflow uses this node and it's still working, you don't need to do anything. If you're setting up the pack fresh: ComfyUI Manager, search "RyanOnTheInside," or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/ryanontheinside/ComfyUI_RyanOnTheInside
cd ComfyUI_RyanOnTheInside
pip install -r requirements.txt

then restart. Both the old and new blend nodes ship in the same install - there's no separate package for the replacement.

Common issues & troubleshooting

It errors or disappears after updating the pack. Deprecated nodes in actively-developed packs are the first things to get quietly dropped in a future release once the maintainer decides the migration window is over. If this node stops loading after an update, that's expected, not a bug - swap the workflow over to Flex Latent Blend.

You're looking for reactive/audio-driven latent blending and this node doesn't seem to have the controls for it. Correct - it doesn't. Despite the name, there's no audio input and no feature modulation on this node at all; whatever audio-reactive latent blending you're picturing is what the replacement, Flex Latent Blend, is actually designed to do.

Manager update stuck. As with the rest of the pack: fully uninstalling and reinstalling the node pack is the author's documented fix if ComfyUI Manager stops tracking updates correctly.