AI-assisted project. This codebase was created with Claude Code. Every figure in it is read from published vendor documentation and cited on the device that uses it, and a test fails the build if anything claims to be
documentedwithout a citation. But none of it has been verified against hardware, and no show has ever been built from a topology this tool proposed — treat a "fits" as a shortlist, not a purchase order. See The honest bit.
Live at willmyshowfit.com — nothing to install.
Describe a show — screens, layers, sources, destinations — and find out which video switchers will actually take it. Not "does the input count add up", but whether there is a physical plug of the right type and capability for every signal, whether the layer budget covers the layers, and whether the canvas fits the chassis. Then it proposes a wiring topology for each switcher that fits.
Nothing is uploaded. There is no backend to upload it to.
Plugs, as a matching rather than a count. Twelve inputs and twelve sources is not an answer: four SDI cameras and two SDI plugs is a "no", and a count will never see it. Every signal is matched to a specific connector, respecting capability — a 4K60 source does not go into an HDMI 1.4a plug, 4:4:4 does not go down SDI, and a DisplayPort link budget is checked as well as its pixel clock.
The plugs that are not what they look like. A Midra 4K output is an HDMI socket and a 12G-SDI socket carrying the same picture: that is one output, not two. Its inputs 1 and 2 are HDMI or 3G-SDI, pick one. Counting the sockets on the back panel overstates that chassis by a third.
The wide-screen layer penalty. Analog Way quotes its layer counts "depending on the screens setup", and this is what that hides: a LivePremier scaling engine spans four output links, so a layer on a screen wider than four outputs takes a second layer link and costs twice. It is a cliff at five outputs, not a slope — going from a four-output blend to a five-output one doubles that screen's entire layer bill, which is how a show exhausts a chassis whose headline layer count looked ample.
Layer capacity in each vendor's own arithmetic. Analog Way charges a 4K mixing layer twice what it charges a DL/2K one. Barco's ladder is 4K : DL : 2K = 1 : 2 : 4. PixelHue budgets layers per output card, so an F8 advertising sixteen 4K layers still cannot put three on a screen living on one card. These are not modelled or derived — they are each vendor's published numbers, cited.
Card capacity, which is not the connector count. This is the limit that catches people out. A Barco Event Master Gen 1 card has four connectors and takes exactly one 4K60 signal between them — Barco's own words are "1 4K60p or 4 HD" — so a single 4K60 source consumes the whole card and strands its other three sockets. An E2 Gen 1 shows 32 input connectors and has eight 4K60 inputs. An E2 Gen 2 shows twelve HDMI 2.0 inputs and can carry 4K60 on eight of them, because they sit on cards rated two apiece.
The backplane, separately. The chassis caps below the sum of its cards: an E2 Gen 1 has four output cards rated one 4K60 each and is still limited to three 4K outputs. Both limits are modelled, and a multi-cable 4K60 counts as one signal, not as its cables.
Canvas and throughput. Barco's live effects canvas in megapixels, with the PGM-only trade-off offered when the show only fits that way. Analog Way's program throughput figure.
Dual- and quad-cable signals. A 4K60 feed on two cables is two half-rate signals, which is exactly why it fits plugs that 4K60 does not.
Passive adapters, where one genuinely works — HDMI↔DVI-D is the same TMDS signalling — and it says so in the patch list rather than pretending the plug matched.
A global toggle. Off, every aux is a plain scaled feed. On, auxes can carry their own layers, and the devices diverge sharply:
| Analog Way LivePremier | Free. Documented outright: layers on aux use adjacent unused outputs, not the main layer budget. |
| Barco Event Master, PixelHue | Charged at the same rate as an on-screen layer. |
| Blackmagic ATEM | Impossible. An aux is a clean routed bus; keys live on an M/E. |
That last row is the point of the toggle: it turns a "which is cheapest" list into a "which can do this at all" list.
The database holds both, in two separate sections, because they answer different questions. A presentation switcher builds an arbitrary canvas across several outputs and edge-blends the joins. A vision mixer does not: every output is one raster, and layers are keyers over a program bus. A 7680×2160 blend is not "too big" for an ATEM the way it is too big for a Midra — it is the wrong shape, and the tool says so in those words.
They are here because a great many shows people reach for a presentation switcher to solve are one screen, one raster and two keys.
Analog Way — Midra 4K (QuickVu, QuickMatrix, Pulse, Eikos, each mode checked separately), Alta 4K (Zenith 100/200), LivePremier (Aquilon RS alpha, RS1–RS6), and the two discontinued lines that are still all over the rental market: Midra pre-4K (QuickVu 3G, QuickMatriX, Pulse², SmartMatriX², Saphyr, Eikos²) and LiveCore (NeXtage 08/16 - 4K, Ascender 16/32/48 - 4K, SmartMatriX Ultra). Barco — S3 standalone, E2 Gen 1, E2 Gen 2, ENCORE3. PixelHue — F4, F8. Roland — V-160HD, VR-400UHD, V-600UHD. Blackmagic Design — ATEM 1/2/4 M/E Constellation 4K.
Card-based chassis are profiled against their stock loadout and, where the chassis misses, searched for a card loadout that would take the show — see below.
Two things the older lines make visible that the current ones do not. A pre-4K Midra has no edge blending at all — soft edge arrived with Midra 4K — so a screen there has to fit on one output, which is a different answer from "you have run out of plugs" and is reported as one. And the whole Midra platform is a 165 MHz single-link machine: 1080p60 and 1920x1200 fit, 2560x1600 and anything 4K do not, in any mode. LiveCore stops at 4K30; to check a 4K60 display against one, describe it as a destination fed by four cables and the tool sizes the quadrants.
Switchers shown sits above the verdict list: pick by brand or by model, with a one-click "only what fits". It is the same selection the report uses, so narrowing the list to two brands while you compare carries through rather than having to be done twice.
The full report is every selected switcher with a wiring topology under each one that fits, which is right when you are choosing a machine and far too much when you are sending three pages to a client who has already chosen. So the report page carries the same switcher picker plus five sections you can drop: the input list, output list, screen breakdown, wiring topology and the source-document links. The example show goes from about 30,000 characters to about 5,600.
Three things stay in whatever you choose. A trimmed report prints a line saying so — "showing 9 of 37 switchers … and it leaves out the source documents" — because a matrix of nine machines otherwise reads as the whole field to whoever receives it. And the caveat stays, because a report that says what fits has to say what that rests on. And every report is signed at the foot — generated by willmyshowfit.com, a free tool from Stoatworks Labs, with the build version, and the line saying it is not a product of any manufacturer it names.
When a modular chassis does not fit as it ships, "does not fit" is only half an answer. The question anyone specifying a build actually has is whether a different set of cards would take it, so the tool searches for one and shows the smallest arrangement that does, with the wiring topology that follows from it.
Loadouts are enumerated in increasing card count, so the first feasible one is minimal by construction; among equals, fewer distinct card types wins (four identical cards are easier to order and swap on site than four different ones), then the tightest fit. The chosen loadout goes through exactly the same evaluation as a stock profile — there is no second, looser notion of "fits".
The search is capability-aware rather than a plug count, in both directions:
- Five 1080p60 HDMI sources get one 8-plug HDMI 1.4 card, not two 4-plug HDMI 2.0 ones, because the cheaper card carries the format.
- One HDMI source gets a 4-connector HDMI card, not the 6-connector multi-format card that happens to have an HDMI plug on it.
- Six 2560x1600 outputs on an E2 Gen 1 get two Gen 2 HDMI 2.0 cards rather than three Gen 1 ones, because a Gen 1 card has only two connectors fast enough (see below) while a Gen 2 card has four.
Slot-position rules are not modelled. Some vendors require particular cards in particular slots, and Barco's Tri-combo cards are capped at 2x 4K60 regardless of having six connectors — so a proposal is a starting point to check against the vendor's own slot diagram, not an order form.
Barco's Gen 1 HDMI output card has four connectors, and the top two run to 297 MPix/s while the bottom two stop at 165. So an E2 Gen 1's eight HDMI outputs are really four that carry 2560x1600 and four that do not — and a show wanting six of them needs a different card, not just a spare slot.
The chassis spec sheets quote only the 297 figure. The per-connector split is published on a different sheet entirely, for a different product. This tool had the optimistic reading until a loadout test disagreed with it.
XML in and out — our own format, attributes a person can read and hand-edit, round-trip pinned by tests. A bad file imports as far as it can and lists what it could not read, rather than throwing away forty good entries over one typo.
Reports as a printable page (browser "save as PDF") or a standalone HTML file, carrying the input list, output list, screen breakdown, compatibility matrix, per-device wiring topology, and every citation behind the verdict.
Every figure here is read from published vendor documentation and cited on the device that uses it. None of it has been verified against hardware. No show has been built from a topology this tool proposed. Vendors revise spec sheets without notice, several of the relevant numbers are published as ranges or with conditions, and a few are internally inconsistent — Barco's E2 Gen 1 sheet duplicates its Aux text into the Mixers row, and PixelHue's F8 sheet disagrees with itself about output slot count. Both are recorded in the data rather than smoothed over.
Figures that could not be sourced are marked unverified and say so in the UI;
a test fails the build if anything claims to be documented without a citation.
Treat a "fits" as a shortlist, not a purchase order.
npm install
npm run devnpm testnpm run build && npm run serve:distnpm run preview does not apply public/_headers, so use serve:dist to test
the CSP for real. See AGENTS.md for the architecture and for what
is verified versus assumed.
Not affiliated with, endorsed by, or a product of Analog Way, Barco, PixelHue, Roland or Blackmagic Design. All product names, logos and brands are the property of their respective owners and are used here only to identify the equipment being described.
MIT — see LICENSE.