Float List To Flex Feature | Akatz

Float List To Flex Feature takes a plain per-frame float list and packages it into the FEATURE type - the custom object (frame count plus a get_value_at_frame(i) method) that some nodes want as input. It's the "pack" half of this pack's FEATURE bridge, and the mirror of Flex Feature To Float List.

You'll want this in exactly one situation: you've computed a nice float schedule - audio amplitudes, brightness, a keyframe output - and the next node in the chain accepts FEATURE rather than a bare list. The node needs a reference original_feature to inherit frame semantics from, and it copies your list's values into that feature shape, so downstream nodes can query per-frame values through the feature interface. It's glue, plain and simple, and it's only worth reaching for when something upstream or downstream is FEATURE-native.

How it works

Under the hood it converts the float list to a NumPy array, takes the length as the frame count, and returns a FEATURE object whose values come from your list - built on top of the original_feature you fed in. Because it reuses the original feature as the container, the two float lists / features in a chain stay consistent about how many frames exist and how values are queried.

Inputs and output

• float_list - the values to wrap, as a FLOAT list (forced input).
• original_feature - a FEATURE used as the container/template; the output inherits its frame-query interface.
• Output is a FEATURE object.

If this feels abstract, that's because it is - it's purely a type/format conversion. In practice it sits in one spot of a workflow: right where a FEATURE-consuming node meets the float-list world.

Installing it

Part of akatz-ai/ComfyUI-AKatz-Nodes. Install via ComfyUI Manager (search "AKatz") or:

cd ComfyUI/custom_nodes
git clone https://github.com/akatz-ai/ComfyUI-AKatz-Nodes
cd ComfyUI-AKatz-Nodes
pip install -r requirements.txt

Restart ComfyUI. No model downloads; requirements are numpy, torch, opencv-python and pydub.

Common issues

The main trap is a length mismatch: the feature's frame count comes from your list, but if the downstream consumer assumes a different total (say, your video's frame count), you'll get out-of-range queries or a schedule that ends early. Match the list length to the target frame count before wrapping. Also, this node is strictly for FEATURE-native chains - if nothing in your workflow actually requires FEATURE, keep everything as float lists and skip the round-trip entirely; you'll have fewer moving parts and nothing to debug.