Accessibility
The toolkit is built to be worked by ear and by keyboard, not only by eye. This page is the whole non-visual manual: voice mode (the site speaks and listens), the vision modes (black sheets, heavy light lines), speech speed, and the keyboard. Everything here also works alongside a screen reader — every spoken confirmation lands in a polite live region too.
Voice mode — the site speaks and listens
The VOICE button sits in the lower left of every page, next to the theme toggle — or press Alt+V anywhere. On: the site reads confirmations and answers aloud and, in Chrome or Edge, opens the microphone and listens for the commands below (other browsers speak but cannot listen — the keyboard still works). The choice persists, so you turn it on once. While the site is talking the microphone is suspended, so it never hears its own voice.
What the mic last heard is printed beside the button, so a sighted deskmate can see what the tool thought you said. Anything that matches nothing is ignored — voice mode does not narrate the room's conversation back at it.
You can always talk over it. While the site is reading, the microphone stays open for the interrupt words: “pause” (also “hold on”, “wait”), “stop”, “resume” / “continue”, “faster” / “slower” (the reading changes speed on the spot), “skip” / “back” to move by a stretch, and “repeat”. Long reads are spoken in sentence-sized runs, so pause and resume land exactly where you left off. Only those exact words act mid-read — everything else the mic hears while the voice is talking is ignored, so the site's own voice through the speakers can never trigger a command.
What you can say — anywhere
| Say | What happens |
|---|---|
| “dark mode” | switch the whole site to dark |
| “light mode” | switch the whole site to light |
| “contrast mode” | black sheets with heavy white lines in every viewport (and dark mode) — the low-vision view |
| “yellow mode” | black sheets with heavy yellow lines — lower glare than white |
| “standard view” | viewports back to white sheets |
| “hover notes off” | stop the source and method cards opening on hover — click a number to read one (screen readers still hear every source) |
| “hover notes on” | source and method cards open again on hover and focus |
| “open surveyor” | “open” plus a tool's name goes there — open studio, open crit room… |
| “go back” | back to the previous page |
| “scroll down” | scroll the page (also scroll up, top of page, bottom of page) |
| “where am I” | hear which page you are on |
| “read the page” | hear this page read aloud — interrupt it any time with pause, stop, faster, skip, or back |
| “pause” | pause the reading mid-sentence (also “hold on”, “wait”) |
| “resume” | carry on from where it paused (also “continue”, “keep reading”) |
| “skip” | jump the reading ahead a stretch; “back” steps it back one |
| “set span to 40” | “set” plus a slider's name and a number moves that slider — any tool |
| “increase span” | nudge a slider up (also decrease, turn up, turn down) |
| “click export” | “click” or “press” plus a button's name pushes it |
| “stop” | stop the voice mid-sentence |
| “repeat” | hear the last thing again |
| “faster” | step the voice speed up (also slower) — each step is a real jump |
| “set speed to 4” | jump straight to a JAWS-class rate — 1 is normal, 6 is the ceiling (also “top speed”, “normal speed”) |
| “help” | hear this list |
| “voice off” | stop listening and go quiet |
The slider and button verbs (“set … to …”, “click …”) work by accessible name — the same label a screen reader announces — in every tool, including the embedded ones. If voice can't find a control, that control is mislabelled for screen readers too; tell us (below) and it gets fixed for both.
Speech speed — up to JAWS pace
The voice runs from 0.6× to 6×. Say “faster” or “slower” to step (each step is a real jump, like a screen reader's rate dial), “set speed to 4” to go straight there, “top speed” for the ceiling, “normal speed” for 1×. The setting persists — an experienced screen-reader listener sets it once and lives there. Above about 2× the site switches to a local system voice, because the browser's network voices silently cap their rate.
Vision modes — black sheets, heavy lines
The VIEW button (lower left, beside VOICE) cycles the low-vision vision modes. They re-render every drawing surface — the sheets that stay white paper in light and dark mode alike — for a low-vision read: inverted to a black background, with the linework thickened and, in yellow mode, tinted. The rest of the page follows the ordinary light/dark theme, and the dark-sheet modes switch the theme dark with them so the chrome matches.
| Mode | Say | What it does |
|---|---|---|
| Standard | “standard view” | Standard view. Drawings back to white sheets. |
| Contrast | “contrast mode” | Contrast view. Black sheets, heavy white lines, dark mode on. |
| Yellow | “yellow mode” | Yellow view. Black sheets, heavy yellow lines, dark mode on. |
Photographs, live camera feeds, colour lessons and the map stay un-inverted on purpose — an inverted photograph is a negative, not a clearer photograph. More modes can be added as they are specified; each is one entry in the mode table.
Hover notes — the source cards, on or off
Every number, layer and stat in the toolkit says where it came from: hover or tab to it and a card opens with the method and the data source. Hover notes (in the Accessibility row of the left toolbar, beside the light/dark switch — or the lower left of a page without a toolbar) turns those pop-up cards off when they get in the way, and back on. Say “hover notes off” or “hover notes on”. The choice persists.
With the notes off, nothing is lost: where a card also opens on a click — the dotted numbers in Proving Ground, for one — a click or tap still opens it; every tool's Data sources and Under the hood tabs are untouched; and a screen reader still hears every source as part of the control's name. The switch only stops the cards jumping out on a mere hover or tab.
The RAP tools — each fully drivable by voice
RAP Studio: every console command works spoken — the grammar is one language for typing, agents and speech. Highlights:
| Say | What happens |
|---|---|
| “describe” | hear the whole model, macro to micro |
| “audit” | hear the spoken ADA/egress audit |
| “acoustics” | hear a room's reverberation (also: acoustics detail, acoustics send) |
| “area site” | square footage (also: area bay a, count columns, extent) |
| “where a” | locate an element — “where”, then its name |
| “add bay b at 30 0” | any console command works spoken — the grammar is one language |
| “stair add from 0 to 1 at 10 10 dir n width 4” | a stair between levels — rise, risers and run derive from the levels |
| “ramp add from 0 to 1 at 10 10 dir e width 4 slope 12” | a ramp — slope 12 means 1 in 12, the ADA maximum |
| “diff” | hear what changed since your mark or the session start (also: diff since import, diff handoff) |
| “undo” | step back one change (also redo) |
Tactile Drawings: “read the drawing” speaks the whole read-back; “plan view”, “section view”, “elevation view”, “axon view” switch the cut; “next sheet” / “previous sheet” step a tiled plate. Every slider in its rail answers to “set … to …”.
Tactile Image: “read the plate” speaks the conversion read-back; the converter itself re-runs on every change, so there is nothing else to trigger.
Texture Compiler (/rap/texture) is meant to be driven with a rig in one hand: load the raking-light frames, say “compile”, and hear the relief depth and coverage read back. One step cannot be spoken and that is a browser rule, not an oversight — a file picker only opens from a real gesture, so “upload photos” presses the button and the operating system’s own dialogue takes over. Everything after that is speech. Say “exposure was not locked” if you shot on an iPhone; no iOS browser can lock exposure, and the compiler solves for the drift instead.
| Say | What happens |
|---|---|
| “upload photos” | opens the file picker to add captured frames |
| “compile” | solves the surface from the frames you have loaded |
| “show height” | shows the recovered relief |
| “show normals” | shows the surface normal map |
| “show albedo” | shows the flat colour, with the lighting removed |
| “show photo” | shows the source frame you are on |
| “next frame” | moves to the next loaded frame |
| “previous frame” | moves to the previous loaded frame |
| “set index to 4” | tells the compiler which detent this frame was shot at |
| “exposure was locked” | best accuracy, if you locked exposure in the camera |
| “exposure was not locked” | solves for the drift instead — use this on an iPhone |
| “download height” | saves the height map |
| “download normals” | saves the normal map |
| “download the tile” | saves a printable STL of the surface, in millimetres |
| “describe the result” | reads out the relief depth and coverage |
| “fix the numbering” | shifts every frame's detent by the offset the photographs support |
| “how many frames” | reads out how many frames are loaded |
| “clear frames” | removes every loaded frame and starts over |
Alt Text has the deepest grammar — describing, band-by-band reading, revising, the camera. It works from the site microphone or its own; all 28 phrases are listed in the tool's voice panel, with these keyboard equivalents. On a phone, Quick Describe (/rap/describe) is its one-thumb shutter: point the camera, one tap, hear what it is — “take the photo”, “read more”, “new photo” by voice.
- D — describe the current image
- A — hear the alt attribute
- 1 / 2 / 3 — hear Macro / Meso / Micro
- E — hear everything
- C — hear the checks
- Q — go to the question box, to ask about this image
- N / P — next / previous image
- O — open or close the camera
- T — take the photo, while the camera is open
- S or Escape — stop the voice
- V — start or stop the microphone
Caustic and the Acoustician — the mixing board by voice
Caustic (the curved-room listening tool) and the Acoustician share one mixing board with eight channels, four groups and a master, and one spoken grammar for it. Say what you mean the way you would to an engineer at the desk: a channel or group name and a control, a number where one matters. The page says back what it did, and for a level it says what the level was before. A phrase it is not sure of is left to the rest of the site; where the site would otherwise move the wrong slider by a guess (“raise the tenor”), Caustic says how to say it instead. The tables below render the real grammar, so they cannot say something it does not do. Whether any particular speech recogniser hears “minus”, or “bass” against “base”, is untested.
| Say | What happens |
|---|---|
| “tenor fader minus six” | Sets the tenor's fader to minus 6 decibels and says what it was before; any channel, group or the master works the same way. |
| “set the tenor fader to minus six” | The same thing, said as a sentence. Only a channel, group or master plus a control name is claimed here, so other sliders still answer to the site's own set verb. |
| “tenor minus six” | The short way: a name and a signed number set that fader. It needs the plus or minus (or zero), so a dropped minus is never guessed; women minus two does the group, master minus eight the master. |
| “bring up the basses by three” | Raises the basses' fader 3 decibels; bring them down, or say altos up three, altos down two. |
| “altos up” | Nudges the altos' fader up 1 decibel; up or down with no amount means 1. |
| “master minus eight” | Sets the master fader to minus 8 decibels; without the word fader the master needs a sign, so a dropped minus can never turn it up. |
| “mute altos” | Mutes the altos; unmute altos brings them back, and it works on a group such as the women. |
| “mute the altos and tenors” | Mutes several at once: up to four names, joined by and or by nothing; solo, unsolo and unmute work the same way. A name it does not know refuses the whole phrase. |
| “solo tenor” | Solos the tenor so nothing else is heard; unsolo tenor releases it. |
| “solo the women” | Solos every channel in the group, because a group has no solo of its own; unsolo the women releases them. |
| “clear solos” | Releases every solo at once, the fix when the mix has gone mysteriously quiet. |
| “pan sopranos left” | Pans the sopranos' direct signal; left or right alone is all the way, or give an amount. |
| “pan sopranos thirty percent right” | Pans the sopranos thirty percent to the right. |
| “centre the basses” | Puts the basses' pan back in the centre. To stand them in the middle of the room, say put the bass in the centre. |
| “flip polarity on the tenor” | Inverts the tenor's polarity, or back again; invert the tenor does the same. |
| “high pass the basses at eighty” | Sets the basses' high-pass to 80 hertz and switches it on; high pass on the basses alone just switches it on. |
| “boost the tenor high mid by three at two point five k” | Sets the tenor's high-mid band to plus 3 decibels at 2.5 kilohertz; cut works the same, and the bands are low shelf, low mid, high mid, high shelf. |
| “women low mid minus three” | A group name with a channel control does that to every channel in the group: pan the men left, high pass the women at eighty, compress the men. Not polarity flips, pings or moves. |
| “compress the soprano” | Switches the soprano's compressor on; compressor off on the bass switches one off. |
| “send the altos to group one” | Routes the altos to group 1, or to no group with send the altos to master. |
| “reset the tenor fader” | Puts that control back at its default. |
| “limiter on” | Turns the safety limiter on. Limiter off is allowed, and says so out loud; please keep it on. |
| “read the board” | Reads every channel's state in a few sentences: what the controls are set to. |
| “read levels” | Reads the meters: the peak of every channel that is sounding and the master, the same words as the Read levels button. It says so when nothing is playing. |
| “what is the soprano fader at” | Says what one control is set to, without moving the cursor: a channel's fader, pan, trim, room or direct send, EQ or compressor value, or the fader of a group or the master. For a switch, the high-pass or a group send, say go to the tenor mute, or read the tenor. |
| “where is the cursor” | Says where the console cursor is: the channel, the control and its value. What control is this does the same. |
| “go to the tenor room send” | Puts the console cursor on that control of that channel, group or master and reads it; nothing on the board changes. Go to tenor alone puts it on the tenor's fader. |
| “read channel three” | Reads one channel or group in detail; how is the tenor does the same. |
| “what's loudest” | Finds the loudest audible channel. |
| “read the room” | Reads the room as the board sees it. |
| “describe the place” | Hears a portrait of the room; what do I hear says the same. |
| “where is everyone” | Says where each singer and the listening seat stand in the room. |
| “where is the soprano” | Says where the soprano stands, and pings from that spot so you can hear the direction. |
| “any echoes” | Says whether the room has flutter echoes or distinct late reflections. |
| “where are the hot spots” | Says where the sound focuses. |
| “what are the numbers” | Reads the room's measured numbers. |
| “what does the seat hear” | Says what the listening seat is hearing. |
| “board help” | Reads the board's own help: the console keys and what they do. |
| “hear it dry” | Plays the mix without the room, for the A and B comparison. |
| “hear the room” | Plays the mix through the room. |
| “room and dry” | Plays the dry mix and the room together, side by side. |
| “flip dry and room” | Switches between dry and in the room; a b does it too. |
| “recall scene two” | Recalls the board you stored as scene 2; scenes run 1 to 4. |
| “store scene three” | Stores the board as scene 3; save scene one does the same. |
| “play the mix” | Starts the choir. Bare play and bare stop belong to the site, so say the mix or the choir. |
| “stop the choir” | Stops the mix; stop the mix works too. |
| “move the sopranos closer by two feet” | Moves the sopranos 2 feet toward the listening seat; farther works too. Distances are in feet. |
| “move the altos three feet left” | Moves the altos 3 feet left on the plan; also right, north, south, east, west. |
| “move the tenor to the wall” | Moves the tenor to the nearest wall; also the centre, the focus, the hot spot or the listener. |
| “put the bass in the centre” | Stands the bass in the middle of the room. |
| “move me to the focus” | Moves the listening seat to the room's focus. |
| “go to seat two” | Listens from seat 2; seats run 1 to 4, and seat three alone works too. |
| “render the room” | Renders the room again after you change its shape. |
| “load the dome” | Loads a room: dome, rotunda, whispering ellipse, torqued ellipse, spiral, snake, arc, parabola, vault, shoebox. For the rotunda say load the rotunda, because open the rotunda goes to the Rotunda tool. |
| “open the torqued ellipse” | Also loads a room; an unknown room name is ignored rather than guessed. |
| “select tenor” | Puts the console cursor on the tenor; channel three alone, or go to channel three, does the same. |
| “mute this channel” | Acts on the console cursor's channel: mute, unmute, solo or unsolo this channel; a group's mute if the cursor is on a group, a refusal on the master. |
| “this channel down three” | Nudges the cursor's fader 3 decibels; up works too, and this channel minus six sets it. The master's fader will not go UP by this phrase: say master up two. |
| “set fader to minus six” | Sets that control on the console cursor's channel, group or master, with no name; a decibel value needs its sign. Room is not claimed this way, because it also names the room's own sliders. |
| “what did you hear” | Says the last thing this page did not take as a command, when you ask. This page never says it on its own, but the site still does for a phrase of four words or fewer that nothing takes: it says I heard, the phrase, and that it is not a command. |
| “where do the first reflections come from” | Says which stretches of wall return the singer's voice to the seat after one bounce, how soon after the direct sound, and how long each stretch is. |
| “who can I hear directly” | Says which channels are heard directly from the seat and which are behind the steel, and the way round to the one you are listening to. |
| “where do the echoes come from” | Says which surface each echo is off, with its share, and whether it arrives from one direction or from all around. |
| “how much of the sound is still a mirror image” | Says how much of the sound that arrives in each window after the direct sound, up to 600 ms, still travels by mirror paths, and how much of the reflected sound in those first 600 ms each window holds. |
| “describe the sky” | Describe the reflection sky: where the reflected sound reaches this seat from, window by window. |
| “compare the channels” | Compares the channels from the seat: which is clearest, which sits furthest below the room, and which cannot be compared. |
| “how far does the focus reach” | Says the nearest way out of the focus, and what the sound does walking from the focus toward your seat. |
| “how does the tone change” | Says how the tone changes with pitch: how long the bass takes to die away against the treble, and how far to trust it. |
| “which bands ring” | Says how long each octave band takes to fall ten decibels, and which of those times come from rays. |
| “compare with the kept room” | Says what you changed in the design and which of the numbers moved by more than the render's own noise, which are too close to tell from it, which could not be compared at all, and what became of each echo; a change smaller than the noise is reported as no change I can tell from rendering noise. |
| “what changed” | Says what has changed on the board since it was opened, a room or project was loaded, a scene was stored, you marked it, or just before a scene was recalled. |
| “mark the mix” | Makes the board as it is now the one that what changed counts from. |
| “undo that” | Undoes the last change to the mixing board and says what it put back. It does not undo the room or where anyone stands. |
| “redo that” | Does again what was just undone on the mixing board, and says what it did. |
| “keep this room” | Keeps the room as it is now, to compare against by ear: change the design, then flip between the kept room and the new one. One room is kept at a time and it lasts only until the page closes. |
| “unpin the room” | Lets go of the kept room, and says so. If the kept room is the one you are hearing, the new room comes back. |
| “hear the kept room” | Switches what you hear to the kept room, its level matched by sound energy, with a short low pair of notes. |
| “hear the new room” | Switches what you hear back to the room you have just been working on, with a short high pair of notes. |
| “flip kept and new” | Flips between the kept room and the new one, each time saying which one you are hearing in a few words. |
The Acoustician answers to the same board phrases with its speakers' names in place of the choir's; what only a room page can do (rooms, seats, moving people, undo) is Caustic's.
| Say | What happens |
|---|---|
| “play the mix” | Starts every speaker playing through the room on the mixing board. Bare play and bare stop belong to the site, so say the mix. |
| “stop the mix” | Stops the board. |
| “solo speaker one” | Solos speaker 1 so nothing else is heard; unsolo speaker one releases it. |
| “clear solos” | Releases every solo at once. |
| “mute speaker two” | Mutes speaker 2; unmute speaker two brings it back. |
| “speaker one fader minus six” | Sets speaker 1's fader — its level in the mix — to minus 6 decibels. The fader is the same number as the speaker's level under Simulate. |
| “bring up speaker two by three” | Raises speaker 2's fader 3 decibels; bring it down works the same way. |
| “pan speaker one left” | Pans speaker 1's DRY signal; in the room, where the speaker stands is what places it. |
| “high pass speaker one at eighty” | Sets speaker 1's high-pass filter to 80 hertz and switches it on. |
| “boost speaker two high mid by three at two point five k” | An EQ move on speaker 2: plus 3 decibels at 2.5 kilohertz. |
| “hear it dry” | Plays the speakers without the room, for the A and B comparison. |
| “hear the room” | Plays the speakers through the room again. |
| “flip dry and room” | Switches between dry and in the room. |
| “master minus eight” | Sets the master fader to minus 8 decibels; the master needs a sign, so a dropped minus can never turn it up. |
| “limiter on” | Turns the safety limiter on. Limiter off is allowed, and asks you to say it twice. |
| “store scene one” | Stores the board as scene 1; recall scene one brings it back. |
| “mute speaker one and speaker two” | Mutes several speakers at once: up to four, joined by and; solo, unsolo and unmute work the same way, and a name it does not know refuses the whole phrase. |
| “speaker one minus six” | The short way to set a speaker's fader: a name and a signed number. It needs the plus or minus, so a dropped minus is never guessed; the confirmation says what the fader was. |
| “read the board” | Reads every speaker's channel in a few sentences: what the controls are set to. |
| “read levels” | Reads the meters: the peak of every speaker that is sounding and the master, the same words as the Read levels button. It says so when nothing is playing. |
| “what is speaker one fader at” | Says what one control is set to, without moving the cursor. |
| “where is the cursor” | Says where the console cursor is: the speaker, the control and its value. |
| “go to speaker one fader” | Puts the console cursor on that control and reads it; nothing on the board changes. |
| “read channel two” | Reads one channel in detail. |
| “what is loudest” | Finds the loudest audible speaker. |
| “what are the numbers” | Reads the traced numbers for the selected speaker at the listener you hear from: reverberation, clarity, strength, envelopment. |
| “any echoes” | Says whether any speaker throws a discrete late echo at the listener you hear from. |
| “where is everyone” | Says where each speaker and each listener stand, in feet, and which way each listener faces. |
| “render the room” | Traces the room again, the same as Run in the toolbar. |
| “select speaker three” | Puts the console cursor on speaker 3. |
Keyboard, screen readers, and what's still rough
Alt+V toggles voice mode anywhere. Escape always stops the voice mid-sentence, even from inside a text box. A skip-to-content link is the first tab stop on every page, every focusable element has a visible focus ring, and every slider's number is a real text input you can tab to and type into.
Honestly still rough: some older forms label their inputs by placeholder only (being swept), and the embedded standalone tools' voice reach depends on their own labelling. If something doesn't read, doesn't speak, or can't be reached without a mouse, tell us — it is a bug, not a preference.