Ino Random Noise

If you've only ever used the plain KSampler, this node looks mysterious - a noise_seed in, a NOISE and a seed out, no image anywhere. It's part of the advanced sampler family: in ComfyUI's custom sampler flow, the sampler takes its starting noise from a dedicated NOISE input rather than a hardcoded seed field. Ino Random Noise is a drop-in replacement for the stock RandomNoise node with one genuinely useful addition: it outputs the seed it actually used.

That second output is the reason to reach for it. The stock noise node swallows the seed; here you can log it, display it, or feed it into a text node so your output filename carries the seed that produced it. Replicating a run goes from "did I save the seed?" to "read it off the node."

How it works

It wraps Noise_RandomNoise from comfy_extras.nodes_custom_sampler - the same generator stock ComfyUI uses - and adds a precision step. The precision toggle is labeled 32-bit / 64-bit and it's a bitmask operation on the seed: on, the seed is masked to 64 bits (& 0xFFFFFFFFFFFFFFFF), off, it's masked to 32 bits (& 0xFFFFFFFF). Then the node hands you both the noise object and the masked seed.

Why would you care about 32 vs 64 bit? Mostly reproducibility and compatibility. Old workflows and some third-party samplers assume 32-bit seeds, and if you're chaining noise from a value you computed elsewhere, the mask guarantees the seed lands in the range the downstream sampler expects. For day-to-day use, leave it on 64-bit; flip it when something downstream complains about seed ranges.

Inputs and outputs that matter

• noise_seed - the seed input, with a control-after-generate option so it can randomize like a KSampler's seed.
• precision - the 32/64-bit toggle described above.

Outputs: noise (NOISE type - wire into a sampler's noise input) and seed (INT - the exact masked seed used, for logging or reproduction).

Installing it

ComfyUI Manager → search "ComfyUI Ino Nodes" → install → restart. Or:

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

Restart after. The pack wants a current ComfyUI (V3-style node API) and installs inopyutils via requirements.txt. No keys, no models, no downloads.

Common issues

The most common mistake is expecting this to replace a KSampler's seed field. It doesn't - it feeds a sampler that accepts a NOISE input (the advanced/custom sampler nodes). If you wire the noise output into a normal KSampler, it'll complain about type mismatch; that's not the node being broken, that's the wrong input. Also worth knowing: this is a fresh noise source, not a noise reshaping node. If your workflow needs to take noise from an existing latent (img2img-style denoise control), this isn't that tool - keep using the KSampler's denoise or the proper sampler chain.