AI Video Estimate

Most "video estimates" still come down to a person watching a call or a clip and typing items into a cube sheet. This is different. The customer films each room from their phone, and the platform detects, measures and prices every item automatically — then hands your manager a reviewable inventory, not a video to sit through.

How it works

The customer opens a link from a text message — no app, no login — and films the rooms. From there the platform runs a multi-model pipeline on the footage:

• Detects the objects in each room and measures their real dimensions using metric depth estimation.
• Self-calibrates against objects of known size in the frame — a doorway, a fridge, an outlet — instead of guessing from a single constant.
• Verifies each item with a vision-language model that confirms the category and reads its color, material and condition.
• De-duplicates the same couch seen across dozens of frames into one item, then matches it to a catalog of 98 reference pieces with real dimensions, cubic volume and weight.
• Prices the move from the measured inventory, packing and complexity — and drops a thumbnail on every item so your manager can confirm the list in seconds.

Every room video has a hard 10-minute analysis budget, so quotes come back the same session — not the next day.

What you get

• A four-model AI pipeline — detection, metric depth, vision-language verification and 3D reconstruction — not a single classifier guessing labels.
• Measured dimensions, not a cube sheet. Volume comes from each specific item's size, not an average for the room.
• A 98-item furniture catalog across 10 categories, each with min/avg/max dimensions, weight and move complexity flags (fragile, needs disassembly, special handling).
• Item attributes read automatically — color, material and condition, the details an estimator normally writes by hand.
• A thumbnail on every item, so a manager reviews the inventory in about 20 seconds instead of re-watching the video.
• Device-aware capture. An iPhone and an Android flagship have different camera geometry — the platform measures against each one correctly rather than assuming one lens.
• Deterministic pricing. The same video produces the same number every time, so you can lock a price before you send it. No "manager A quoted $1,400, manager B quoted $2,100."
• The inventory flows into the contract. The AI item list pre-fills the DOT Bill of Lading — no re-typing between the estimate and the paperwork.

Why it beats a plain "video estimate"

Most video-estimate tools on the market are a video call or a recording that a person still has to eyeball and count. The inventory is built by hand; the volume is an average from a table. Here the objects are detected, measured and priced automatically, the estimate is repeatable, and the same data becomes the Bill of Lading — instead of a manual re-entry between the quote and the truck.

The numbers

• 10-minute hard analysis budget per room video — same-session quotes.
• 12-step, 4-model pipeline on every video.
• 98 reference furniture items across 10 categories, with measured dimensions and weight.
• Compared with a typical on-site estimate — an estimator's round trip plus a day of scheduling — this is a link in a text and an answer the same day.