Seed Topology Mapper (Arctenox's Essentials)

One base seed in, up to 16 related seeds out, arranged by math instead of vibes. Seed Topology Mapper takes a single base_seed and generates a deterministic family of seeds using one of six spacing schemes: golden_angle (the ~137.5° golden angle), harmonic (musical ratios 1:2:3:4…), fibonacci, chaos_neighbors (a deterministic logistic map), prime_spiral, and phi_spacing.

Who needs that? Anyone doing batch or variation work who wants the seeds to be reproducible and organized rather than a dice roll. You set a base seed once, the node hands you seed_1 through seed_16, and you feed a different one into each of your KSamplers. Same inputs tomorrow = same seed family tomorrow, which makes sharing and debugging a workflow much easier than "seed: random, pray."

How it works

Pure math, no model, no GPU. Each scheme computes an offset from the slot index and a spread multiplier, then combines it with the base seed (normalized to 32-bit). The README's recommended spread ranges are worth respecting: golden_angle 0.5–2.0, harmonic 1.0–5.0, fibonacci 0.1–1.0, chaos 0.5–1.5, prime_spiral 1.0–3.0, phi_spacing 0.5–2.0.

One honest reality check: mathematically-related seeds are still effectively uncorrelated in latent noise space. Bumping a seed by 1 already gives you a completely different image, so the "aesthetically related variations" framing is optimistic - a golden-angle neighbor won't reliably look like a subtle cousin of the original. What the node genuinely buys you is deterministic, named, shareable seed sets, not guaranteed visual kinship.

The inputs that matter

There are only four, and three of them are the whole node:

• base_seed - where the family starts.
• topology_type - which spacing scheme generates the offsets.
• num_seeds - 1 to 16, how many related seeds you want.
• spread - a spacing multiplier (0.1–10.0). Lower = closer spacing between seeds, higher = more spread.

The outputs

seed_1 through seed_16, all INT. Slots beyond num_seeds output the base seed unchanged - so you can wire all 16 outputs and only the first N actually vary. From there each seed feeds a KSampler's seed input. If you're pairing with this pack's own KSampler (Arctenox's Essentials), grab its seed_used output to log the exact seed each pass actually sampled.

Install

ComfyUI Manager, search Arctenox's Essentials, install, restart. Or:

cd ComfyUI/custom_nodes/
git clone https://github.com/Arctenox/Arctenoxs-Essentials_ComfyUI

Restart. No model downloads - just torch, numpy and optional psutil.

Gotchas

• If all you want is "a bunch of different seeds," a randomizer does that in one node with less ceremony. This earns its place when the seed set itself needs to be reproducible and shareable.
• Don't expect visual similarity between family members; treat the topology as an organizational scheme, not a creative dial. If you genuinely want near-variations, that's denoise or prompt perturbation territory, not seeds.
• The pack is marked DEPRECATED - Arctenox is remastering it into a new combined pack. Works fine today; just don't treat it as permanent infrastructure.