💪TB | Unpack Float

Same family as UnpackBoolean, different data type. 💪TB | Unpack Float reaches into the pack's genparams bundle, reads one named numeric parameter, and outputs it as a plain FLOAT that any standard ComfyUI node can consume.

The BOOLEAN version returns False when a key is missing. This one is a touch friendlier: you give it a fallback value, and that's what comes out when the parameter isn't there. That makes it safe to wire into a graph that should keep working even when a prompt didn't specify the value - which is most of the time, honestly.

The three inputs

• genparams - the bundle to read from.
• param_name - the dotted key to read, e.g. denoising.upscaler.extra_noise.
• default_value - what to return if the key is absent. Default 0, and the step of 0.1 just controls the widget's increment, not what you can type.

Output is a single FLOAT.

What it's for

In the pack's native workflow, floats travel inside genparams - --cfg-shift -1, --image-shift +2, --upscale-noise 0.6 all end up as nested numeric keys. Unpack Float is how you surface one of those to the rest of the graph: feed a float into a standard sampler, a scale node, a math node, or just to see what a value resolved to while debugging. The source is a single genparams.get_float(param_name, default_value) call - nothing clever, which is exactly what you want from a plumbing node.

Pair it with Unpack Boolean and UpackSamplerParams and you can dismantle the genparams bundle into individual, inspectable values - useful when you're mixing TinyBreaker nodes with the broader ecosystem and want the whole pipeline readable on the canvas. If you never mix packs, you may never need it; that's fine. It's there so you can escape the bundle when you need to.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/martin-rizzo/ComfyUI-TinyBreaker

or via ComfyUI Manager ("tinybreaker"), then restart. No pip dependencies. The usual model files apply to the pack as a whole (tinybreaker_prototype1.safetensors, t5xxl_fp8_e4m3fn.safetensors), though this node, like its boolean sibling, reads metadata and doesn't need the models loaded to run.

Remember the dotted-key gotcha: param names must match exactly, and a miss just yields your default. If a value keeps coming back as 0 when you expect something else, you're almost certainly mis-spelling the key or reading a section of genparams that was never written.