✍ Float Input

Float Input [DVB] is a single float value in, a FLOAT out, nothing else. It is the pack's version of the primitive number box - and for a node that does literally nothing, it earns its place in the Dream Video Batches workflows in a way that surprises people. Look at the pack's own example workflows and you'll see these nodes everywhere, titled things like "Seconds fade" and "Seconds per image", feeding durations into transitions and repeats. That's the whole point: named, typed constants instead of anonymous primitive widgets.

Why it exists

The pack's author built a small family of these - Float Input, Int Input, String Input, Text Input - and they're workflow-hygiene nodes. When you wire a bunch of numbers into a graph, a row of bare primitives tells you nothing; a node titled "Fade seconds" tells you everything. The pack's own examples lean on them hard precisely because a complex video pipeline gets unreadable fast, and a titled constant box is the cheapest fix there is.

There's a practical wrinkle too. This pack's transition and math nodes take floats in real units - fade_seconds, factor, fade_seconds again - and having those as named nodes means you can tweak "the fade" in one obvious place instead of hunting through three widget rows. Feed a Float Input into Blended Transition, Fade From/To Black, Frame Set Frame Dimensions Scaled, or Divide and the intent of the graph stays legible.

The honest take

If you're building small graphs, you don't need it. But it costs nothing, it's stable, and when a workflow from this pack's README or examples asks for it, it's the node they mean. The one real distinction to remember is against its sibling Int Input [DVB]: floats for anything fractional like seconds or scale factors, ints for frame counts and indices. Wire a float where the graph expects an int and some nodes will still accept it (the pack's nodes do loose conversion in places), but don't count on it - keep the types straight.

Install

Same as every node in the pack:

cd ComfyUI/custom_nodes
git clone https://github.com/alt-key-project/comfyui-dream-video-batches.git
cd comfyui-dream-video-batches
pip install -r requirements.txt

Restart ComfyUI, or search "Dream Video Batches" in ComfyUI Manager. No models, no heavy dependencies - just the pack's standard pip list (imageio, pilgram, scipy, numpy, torchvision, evalidate). The only real installation gotcha in this pack is the numpy<2.0 pin, which can clash with other packs that want numpy 2.x; if you hit that, resolve the conflict at the environment level rather than removing the pin.