Frankenswitching: building a custom switch out of other switches' parts
Four parts, four independent axes, and a mix-and-match history that predates every factory switch that eventually copied it — the original Holy Panda was a pile of parts from two different switches years before anyone sold it as one SKU.
Open any switch and four parts fall out: a top housing, a bottom housing, a stem, and a spring. Nothing requires those four parts to have started life in the same switch. Frankenswitching is the practice of taking the part that does one thing well from one switch line and marrying it to the part that does something else well from another — and it has been going on long enough, and worked well enough often enough, that the most famous result eventually got sold at retail as its own product with its own part number.
The idea
A switch's feel breaks cleanly into four independent levers, and a builder with a parts bin and a switch opener can pull any one of them loose and swap it. It's the same divide-and-conquer instinct the hobby applies to servicing stabilizers — isolate the one contact point actually causing a problem and fix only that — just run across a switch's four parts instead of a stabilizer's three contact points.
The top housing sets the acoustic shell — the material and wall geometry that the stem and leaf are striking against — and carries part of the pole-fit tolerance. The bottom housing sets the base tolerance: rail length, rail width, and where the stem rests before a keystroke begins. The stem carries the tactile bump profile (or the absence of one, for a linear) and the pole geometry that rides inside both housings. The spring sets actuation weight, bottom-out weight, and the force curve between them — the same three levers covered in detail elsewhere.
None of those four has to agree with the other three about which switch it came from. The canonical proof is the original Holy Panda: an Invokim/Everglide Panda top and bottom housing, with a Halo True (sometimes Halo Clear) stem swapped in for the Panda's own. Neither switch alone produced the combination builders wanted — the Panda housing's deep, rounded shell paired with a tactile bump shape neither Invokim nor Halo shipped on its own. Enough builders did the same swap by hand that the combination earned a name, and years later Drop Holy Panda X shipped as a factory-built switch, available without opening a single housing — the clearest signal the hobby has that a frankenswitch recipe worked.
The parts list
| Component | What it actually controls | What builders usually mix in it for |
|---|---|---|
| Top housing | Acoustic shell tone, leaf geometry, part of the pole-fit tolerance | A quieter or deeper shell material without changing anything else |
| Bottom housing | Base tolerance, rail length/width, resting position of the stem | Tighter tolerance for less wobble, or a specific housing's bottom-out feel |
| Stem | Tactile bump shape (or none, for linears), pole geometry | Borrowing a bump profile or a smoother pole from a switch with a housing nobody likes |
| Spring | Actuation weight, bottom-out weight, force curve | The lightest-touch change — swap without opening anything else on the list |
Why the combination works
The reason frankenswitching is a repeatable technique rather than a lucky one-off is that these four properties genuinely don't interact much. A top housing's acoustic signature doesn't change because a different stem is riding inside it; a stem's bump shape doesn't change because it's seated in an unfamiliar bottom housing. Each part is doing its own job, largely independent of what the other three are made of — the same discipline that makes spring swaps a precision tool rather than a vibes-based one applies here across all four parts at once, not just the spring.
That independence is exactly what the Holy Panda recipe exploited. Builders wanted the Panda housing's rounded, muted shell and did not want the Panda's own stem; they wanted Halo's tactile bump and did not want to give up the housing they already liked the sound of. Mixing solved a problem neither factory switch solved alone.
Lubing is a fifth axis sitting on top of all four, and it's the most independent of all — it changes feel and sound without touching which switch any of the four donor parts came from, so it layers cleanly onto any frankenswitch recipe rather than complicating it.
Building one
The order matters. Pick the host housing first, usually for its acoustic shell — the part hardest to change any other way. Source the donor part next: a stem for its bump shape, a bottom housing for its tolerance, a spring for its weight. Before opening every switch on the bench, test-fit one donor part into one host housing and check for binding or play across a full keypress. Only after that single pairing feels right does the swap scale to a full board's worth of switches.
Foam dropped inside the housing itself is a different axis again. The Holee mod changes acoustics without touching pole fit, tolerance, or which switch any of the four parts originally came from, so it stacks on top of a frankenswitch recipe rather than competing with it.
Caveats
Three things go wrong often enough to name individually.
Stem-pole spacing and wobble. Housings are molded around a specific pole shape and spacing. A stem from a different switch line can seat and still wobble more than either donor switch did alone, because the pole fit is now a compromise between two tolerance specs that were never designed to meet. Test a single switch through full travel by hand before assuming the fit is clean.
Spring length and weight mismatch. A spring's free length is matched to its housing's compression range at the factory. A spring that's too long for the new housing can prevent the stem from resetting cleanly; one that's too short changes how early in the stroke resistance builds, which shows up as a softer or mushier bottom-out than either donor switch had.
Housing tolerance stack-up. Pairing a loose-tolerance bottom housing with a tight-tolerance top (or the reverse) doesn't average out to a medium fit — it produces either a rattly switch or a binding one, depending on which half is tighter. Tolerance doesn't blend across a seam; it stacks.
Variants to try
The safest entry point is a spring-only swap — it touches none of the fit-dependent axes above and is reversible in minutes. Next-safest is a stem swap within the same mold family, where pole geometry is already a close match. Riskier, but where most of the acoustic payoff lives, is a top-housing swap on an otherwise-stock switch, changing shell tone while leaving the bottom housing and stem exactly as they shipped. Builders who've already run through the cheaper acoustic levers — the tape mod chief among them — tend to reach for frankenswitching next, once those easier wins are exhausted.
A hypothetical worth bench-testing rather than buying in bulk: a Durock T1 top housing, chosen for its sharp bump near the top of the stroke, paired with a Gateron Oil King bottom housing and stem, chosen for the deep, factory-lubed voicing Oil King ships with. Nobody has documented that pairing the way the Holy Panda recipe is documented — it is exactly the kind of untested idea this hobby runs on, and it is a single-switch test, not a 90-switch bet.
Full four-way mixes — different top, bottom, stem, and spring all at once — are where the risk compounds fastest, since a bad fit on any one axis is hard to isolate from the others once everything's reassembled. Worth watching: as more vendors follow Drop's lead and productize bench-tested combinations, the trial-and-error that built the Holy Panda recipe gets easier to skip entirely — buy the factory version, or at minimum start from someone else's documented pairing instead of a clean-sheet guess.
Judge the result the way any mod should be judged: over a full day's typing, not a sixty-second check. See why typing tests lie for the gap between a quick benchmark and how a recipe actually holds up on the clock.
Keep reading
- Ideas
Lubing switches: what to use, how much, and what to skip
Four lubes, four application points, and one rule that prevents most mistakes: over-lubing is irreversible, under-lubing is not. Here is the practical reference.
thock10 min read - Ideas
Spring swaps: what swapping the spring actually changes
Swapping the spring is the most-discussed mod that does the least of what builders expect. Here is what the spring actually controls, what it does not, and when the swap is worth the afternoon.
thock9 min read - Ideas
Switch films: do they actually do anything?
Thin POM shims between your switch housings cost five dollars and promise a tighter, better-sounding switch. The mechanism is real. Whether it matters on your specific switches is a separate question.
thock8 min read - Ideas
The Holee mod: what foam inside the switch actually buys you
Small pads of polyurethane foam cut and wedged inside the switch housing itself — not under the PCB, not behind it, not under the keycap. The Holee mod works at a smaller scale than every other foam mod in the hobby, and it costs proportionally more time.
thock8 min read