Flex Feature To Float List | Akatz

Flex Feature To Float List converts a FEATURE object back into a plain list of per-frame floats. It's the "unpack" half of this pack's FEATURE bridge - the thing you reach for when a node hands you a FEATURE and the next node wants a boring old float list.

FEATURE is a custom type this pack (and friends) uses to carry a per-frame value stream with metadata - a frame count and a get_value_at_frame(i) method - instead of a bare list. Some nodes speak FEATURE natively; lots of others don't. When you're in a workflow where the FEATURE comes from one side but the consumer on the other side wants a plain schedule, this node is the adapter. It iterates every frame, pulls each value, and hands back the list.

How it works

Mechanically it's a loop: for i in range over the feature's frame_count, call get_value_at_frame(i) and append. The output values are whatever the feature holds per frame - often normalized 0–1, but it passes through exactly what the feature stores, so the meaning depends on the feature's source. There's no rescaling, no truncation - what you see is what the feature had.

Inputs and output

• feature - a FEATURE-type input (forced input, so it has to come wired from another node).
• Output is a FLOAT list, one value per frame.

That's the whole node. Round-trip note: Float List To Flex Feature is the reverse direction, and because that node rebuilds a feature from a list, the two make a clean list ↔ feature ↔ list loop with nothing lost along the way.

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 realistic gotcha is feeding in a feature from a different pack whose frame_count or value semantics differ from what this pack expects - the node trusts the feature's own interface, so garbage in, garbage out. Also watch the list length against your frame count downstream; a feature of 100 frames produces 100 values even if your target video is 120 frames, so pair it with Adjust List Size if alignment matters. Honestly, for most workflows you'd rather keep values as a plain float list end-to-end; the FEATURE round-trip is for when the source node only speaks FEATURE.