Every entry below is a real job — make the listing film, animate a scene, cut a reel, publish it. They are not eleven products. Each one names the capabilities that serve it, inherits the weakest status among them, and states what it will refuse to do.
4 capabilities live3 next
No film can be rendered in this build. Accounts, teams and the credit ledger are what you can use tonight.
01Next milestone
A listing video maker built around one property
What you need
You have a live listing, a folder of photographs and a portal that wants a video by Friday.
What we will do
One workspace: property facts, ordered photos, an editable scene plan and a cost line, then a cut per aspect ratio.
The maker is deliberately narrow. It takes your photographs as the source of truth and animates only what they contain — a push, a pan, a held frame. A tool that re-imagines a room is not a listing video maker; it is a renderer with a liability attached.
Today: The workspace, roles, invitations and credit ledger are live. The photo upload and scene plan are the next milestone; nothing renders yet.
Never: It will not accept a floor plan and invent a walkthrough, and it will not add a room that was not photographed.
Some listings deserve movement that a still cannot carry — a slow drift across a terrace, light shifting in a room.
What we will do
Cinematic AI is an opt-in per scene. Each clip records the provider, model version, duration and price before it runs.
Fidelity stays the default because it is cheaper, faster and cannot hallucinate a wall. AI motion is a choice you make per scene, not a mode the product slips into, and a rejected shot can be reverted without orphaning the rest of the film.
Today: The provider adapter, queue, budget guard and result verification are built and tested against a simulated provider. No paid generation has been run, and no UI can start one.
Never: It will not generate people, pets or new furniture by default, and it will not call a filmed room an inspection.
The same property has to appear as a 16:9 master, a 9:16 reel, a square post and a 4:5 story.
What we will do
Each ratio is composed from the same scene plan rather than cropped from a master, so doorways and faces stay in frame.
Re-composing per ratio is more work than cropping and it is the difference between a reel that reads and one where the agent's phone number is cut in half. Caption cues are stored as data, so the same cut can carry French, Thai or Arabic text without a re-render of the pixels.
Today: These ratios exist as product decisions and in the schema. Export does not exist, so no ratio can be downloaded.
Never: No format will be advertised before it can be downloaded from the product.
You already paid a photographer. You do not want another shoot; you want the stills to move.
What we will do
Fidelity mode animates your photographs deterministically: crops, pans, zooms and pacing, never invented pixels.
Every scene is a photograph plus a move plus a duration. That means a client review is a review of a plan, not of a model's mood, and a failed job can be retried without re-shooting the property.
Today: The motion model, the credit ledger and the job queue exist in the codebase. Upload, plan and render do not.
Never: It will not claim a still was a video take, and it will not label a motion graphic as footage.
Different agents, one brand: the address, the price, the licence line and the logo must be right on every film.
What we will do
A brand kit per workspace, a price cartouche per language, and facts frozen into the project at generation time.
Freezing the branding snapshot onto the project is what stops a later kit edit from silently changing a film a client already approved. It also makes the six concept identities on the landing page an honest demonstration of a real design decision.
Today: The conceptual identities on the landing page demonstrate the workflow. Saving a brand kit is not built.
Never: No awards, ratings or review counts can be entered as text and rendered as if verified.
A listing film has to earn its keep: a hook, captions that work muted, a call to action and a version per channel.
What we will do
Per-channel compositions from one plan, captions timed as data, and a CTA that survives every crop.
Publishing is specified, not built: the provider scopes, the token storage and the queue are written down before any button is designed. Until that exists, the marketing output is a file you post yourself.
Today: Nothing publishes to a social network. There is no connector, no schedule and no OAuth app — the product says so rather than showing a fake calendar.
Never: No invented reach, no "posted automatically" badge, no scheduled post that will not fire.
You want the film assembled quickly without a model redesigning the property or inventing a feature.
What we will do
AI proposes order, pacing and captions; the images, the price and the contact block stay yours.
The director is the part worth automating: which photograph opens, how long a corner bedroom holds, where the caption sits. The parts worth refusing to automate are the facts — a model that writes the price is a model that can write the wrong one.
Today: The AI Director is a specification. The pipeline that will execute its plan exists as a tested provider layer, not as a product surface.
Never: The AI will not author legal or property facts, and it will not generate a review, a rating or a counter.
Buyers want to feel the sequence of a home before they visit it.
What we will do
A linear, directed sequence through your photographs with consistent motion and honest pacing.
This is the one place where the category lies most often. A film that moves between photographs is a tour in the sense that a brochure is a visit: useful, honest when labelled, and not a dimensionally accurate model. VividProp says "directed film" and never "3D tour".
Today: Not built. There is no scene planner UI and no renderer.
Never: It will never be called a 3D tour, claim measured geometry or offer a navigable model from a linear film.
The first cut is never the one published, and a full re-render for a trimmed scene is a waste.
What we will do
Quick edits re-cut rendered segments; only a replaced scene runs the model again, with its own cost line.
Because a scene is a photograph plus a move, editing stays structural: hold a frame, drop a scene, swap the ending card. That is also why the editor will not accept arbitrary MP4s — it edits plans it can re-execute, not footage it cannot read.
Today: Not built. The editor is specified down to the caption cue and the version record.
Never: It is not a replacement for an NLE, and it will not accept footage it cannot re-render.
Agencies often buy a retainer: an editor, a turnaround, an invoice per listing.
What we will do
VividProp is self-serve software with credits shown before a job runs. No account manager, no per-listing quote, no surprise reel.
That is a positioning decision, not a missing feature. A services model hides the cost inside a conversation; a metered product puts the number on the screen. If you need a human editor for a specific shoot, the honest answer today is that VividProp is not that vendor yet.
Today: Accounts, teams and the credit ledger are live. Plans are not charged during early access, and no service retainer is offered.
Never: We will not advertise a done-for-you retainer we cannot staff.
Your listings already live on a portal; retyping the address and re-uploading twenty photos is waste.
What we will do
Import the listing's own data, keep the source URL on the project, and ask before reusing an image you may not hold rights to.
Scraping a portal is only half the problem: the photographs usually belong to the photographer or the portal, and re-publishing them in a new film is a rights question, not an engineering one. The feature arrives with a rights confirmation step or it does not arrive.
Today: Not built and not scheduled. No portal integration exists, so no import button is offered.
Never: We will not silently re-host photographs whose licence we cannot establish.
The category sells all eleven of these as shipped. The honest version costs us conversions and it is the only version worth building: a status page that stays true after launch is the differentiator.
Statuses are computed from a single capability registry shared with the product, not typed into this page. A tool inherits the weakest status behind it, so nothing here can read as finished while its pipeline is not.