Walk & talk: narrate items while you scan

Walk and talk is the heart of a LiDAR scan. As you walk a room with the scanner running, you say what you see, and each thing you say slides in as a card over the live camera. When your words line up with a piece the scanner just measured, the card carries that real, measured cube. This article goes deep on the narration itself. For starting and finishing a scan, see Scan a room with LiDAR.

You'll need: An iPhone or iPad Pro with a LiDAR scanner, and microphone plus speech permission (the app asks the first time).

Say what you see

Speak naturally, one item or one thought at a time. A short pause (about 1.3 seconds) ends what you're saying and turns it into a card. The mic pill at the top shows the live transcript so you can see it hearing you: "Listening — say what you see" when it's quiet, then your words as you speak.

• Say "oak dresser" and a card appears for it.
• Say "six dining chairs" and you get six pieces (see the quantity rule below).
• Say "scratch on the top left corner" as part of naming a piece and the app turns it into a condition on that card.
• Say "pile of stuff on this counter" and it opens the misc-pile flow.

Tip: The mic keeps listening for the whole scan. You don't tap anything to talk. If another app grabs the microphone (a phone call, Siri), the pill honestly switches to "Voice paused — another app took the microphone" instead of pretending to hear you.

Reading a card: MEASURED vs. EST.

Each card shows the item name, its cube and weight, a category, and a badge:

• MEASURED means LiDAR locked a real measurement onto that piece. The badge is tinted (accent color).
• EST. means the cube is a catalog estimate, not a measurement.

You also feel which way it landed without looking down:

• A firm tap (a rigid haptic) means a LiDAR measurement locked on.
• A soft tap means it landed as a catalog estimate.

VoiceOver announces the same thing, e.g. "Oak dresser, measured 30 cubic feet."

What each card button does

Every card is a suggestion. Nothing joins the room until you act on it.

Only the two newest cards ride the viewfinder at once. If more are waiting, a "+N more waiting" pill appears; tap it (or the tray at the top) to open the session list, where you can add or dismiss any card.

Important: Saying an item does not add it. Every card waits for Add, and the review screen after you finish is the final check. This confirm-first rule is what keeps the inventory truthful.

How the app matches your words to a measurement

The app only puts a measured cube on a card when it's confident the words and the measurement are the same piece. This is deliberately strict, because printing one object's measurement on another object's line would be a wrong legal record. A narrated item claims a measured object only when all of these hold:

1. One piece, one measurement. The utterance names exactly one piece. If you set a quantity above one, the card gets catalog estimates instead (see below).
2. The 10-second window. The measured object was detected or updated within about 10 seconds before you started talking. You describe what you were looking at when you started; something first seen after that is treated as a different thing across the room.
3. The name fits. The spoken name passes a keyword check for the object's coarse RoomPlan type. RoomPlan only knows a handful of broad shapes, so "dresser," "bookcase," "wardrobe," "armoire," "chiffonier," and the like all fit its "storage" type; "sofa," "couch," "loveseat," "sectional" fit its "sofa" type; and so on. A name that fits nothing measurable just lands as an estimate.
4. The object is unclaimed. No other card already took that measurement.

When two detections both qualify, the newest wins; if they're near-simultaneous, the one closest to where the camera is pointing wins. If nothing matches, the item still lands, clearly badged EST. A missed match costs a little precision; a wrong match would be a false record, so the app errs toward the estimate.

Quantity above one drops the measurement

Say "six dining chairs" and you get six pieces, each an estimate. A measurement belongs to exactly one physical piece, so any quantity above one releases the single LiDAR measurement and gives every piece the catalog estimate. If you want each chair measured, scan and name them one at a time.

The same happens in the card editor: raising quantity past one on a measured card shows "More than one piece releases the LiDAR measurement — each piece gets the estimate instead," and drops the measurement.

Note: If you type a cube by hand in the card editor, that becomes your figure and outranks the scanner's measurement. The card re-badges to reflect that it's no longer the raw measurement.

Speaking damage

You can code pre-existing damage with your voice as part of naming a piece. The app maps your own words to the standard condition codes and van-line location symbols, deterministically, from what you actually said. It never guesses damage in.

• "scratch on the top left corner" codes as corner + scratched. When several locations are near a damage word, the most specific part wins (a corner beats a side).
• "dresser with a scratched top and a dented leg" codes the top as scratched and the leg as dented, pairing each damage word with the nearest location word.
• A spoken phrase that matches no standard code stays visible on the card as a heard note so you can code it by hand. It is never forced into a code.

Damage codes decoded from your words appear as chips on the card, and go on the inventory only once you tap Add and keep the item on the review screen. For the full damage system, see How damage codes work.

Photographing damage mid-scan

Tap the camera button on a card to grab a frame from the live scan as a damage photo, no second camera needed. If your company has AI photo analysis on, the app uploads it and suggests defects you can review right on the card; each suggestion is accepted or rejected individually and never self-applies. Any photo that can't upload right then stays attached and uploads when the room is imported.

Misc piles

For loose goods that get boxed rather than moved as-is (a counter of kitchen items, a closet shelf), flag a misc pile: tap Misc pile at the bottom, or say a pile phrase like "pile of stuff on this counter," "counter full of…," or "drawer full of…". The app opens the Pack this pile sheet where you describe the pile, optionally photograph it, get a carton suggestion (if AI is on), adjust the counts, and add them. Confirmed cartons add to the inventory as CP cartons. See Boxes & cartons.

Note: One pile at a time. If a pile sheet is already open and you say another pile phrase, the app tells you "Finish the open pile first — one pile at a time." so your words are never swallowed silently.

The session tray

The pill at the top right shows the room so far: confirmed item count, total cubic feet, and carton count (e.g. "4 · 62 ft³ · 3 ctn"). Tap it for the full session list: items waiting for review, items added this room, and packing cartons. Swipe a waiting card to add or dismiss it; swipe a confirmed one to remove it.

Common mistakes

• Assuming what you said was added. It wasn't until you tapped Add and kept it on the review screen.
• Saying a quantity when you want a measurement. "Two nightstands" gives two estimates. Name single pieces one at a time to measure each.
• Coding damage on a multi-piece card. Codes on a card with quantity above one go on the first numbered piece only; the others import clean. Narrate damaged pieces one at a time to code each.

Related articles

• Scan a room with LiDAR
• The review screen after a scan
• Boxes & cartons
• Record pre-existing damage
• Categories & how weight is estimated