This is part of a series of markets about things I could do in 2024. See also: /Mira/which-of-mira-s-cool-ideas-will-mir
Summary
Real estate agents spend thousands of dollars renting furniture to take photos when listing houses or apartments. What if I run it through a couple AI models, so they can add couches, tables, turn the cloudy sky into a bright sunny blue one instantly?
There is already precedent using CAD. I believe they add 3D geometry to the photos, and then place props of furniture. 24 hour turnaround, $20-80 an image.
A lot of this depends on getting segmentation, Stable Diffusion, and other models to reliably do this. Because it would look bad if someone had to try 20 different generations or prompts before they got something reasonable. Early on, I could match the 24-hour turnaround and generate the images myself to make sure they are reasonable quality. In the mid-term, rerunning prompts also sounds much cheaper to hire for than a 3D CAD person.
Each MLS has its own local rules for reporting this, so my software would need the ability to put a watermark on the image to indicate our name and that it has been "touched up". This toggle can be optional, since it is the agents' responsibility.
Market Mechanics
Trigger condition: Deploying any application website to a domain name. Planning, writing prototype code, a static placeholder website, etc. will not be sufficient as minimal effort for this market.
Resolves NA if the trigger condition is not met. Otherwise, resolves to the logarithmic interpolation between nearest bounding answers(or to $1 if I start this but don't make any positive profit for the entire year). Units are "profit in USD". I'll add new options at multiples of 10 as needed before resolving.
Example: I make $500 in annual profit. The nearest options are $100 and $1000. Then this market resolves (ln(500) - ln(100)) / (ln(1000) - ln(100)) = 70% $1000 and 30% 100.
"Profit" is income minus expenses. I'll have to calculate expenses for my taxes, so e.g. OpEx will count 100% as an expense, CapEx will probably be weighted 20%. A salary to myself would count as an expense. Equity doesn't count as income.
Related questions
Doesn't seem like a winner idea - especially if the focus is on price.
This is because real estate is so expensive that renting actual furniture and a professional photographer would account for only a small part of the price. Which is recouped if the photos increase the selling price by a few percent - saving 80 bucks does not make sense here.
I think for this to work, it should 1/ be truly high quality (it's a conservative industry, usual image generation quirk will not be tolerable when actual photography is a totally viable alternative) 2/ focus on things that would be very hard to do physically. Maybe an option for people looking at the ad to "load their own furniture", or at least to have a large selection of style. Creativity is important here. 3/ Modern tooling, maybe offer an embed instead of static images, access to prompt for ad looker, 3d render, whatever 4/ Don't focus on low price. Charge more but have more quality. Same reasoning as the photo crew: real estate is expensive.
@CamillePerrin $20-80 per image not for the whole set. I would probably charge $500 for a set(0.5% of some houses in the US), and would publicly compare myself to physically renting furniture not to 3D CAD companies. My solution could also work on really cluttered houses(imagine a multifamily or a foreclosure). Hopefully agents have enough volume that they would purchase multiple sets in a year, and then I would only need a couple hundred people or so to reach $500k income, which is similar to the baseline of getting a job, which would be a success.
I might also be able to infer the 3D geometry given enough examples and sell access to that model to the 3D CAD companies themselves, so they need to hire fewer 3D CAD people to add geometry. That would be different than what I'm described, but would count as the same concept.
@Mira Under the (somewhat big) assumption that it works technically, which is made easier if you're only targetting ~200 people a year (allowing unscalable things like the occasional manual intervention), that could work at steady state.
But what's your plan to get your first ~10 customers to start the whole thing? Do you know realtors/have contacts in Zillow or whatever? Doesn't seem like a very tech forward industry, for instance the "3D virtual visit" thingy has only relatively recently become widespread, despite the technology being fairly old (about as old as Google maps "photo sphere" since that is basically what it is)
There is also a legal aspect that could be involved with AI generated image - if a buyer can successfully argue to have detrimentally relied on features of the property that were Ai generated but not actually true (E.g. living room was shown with sun light, but it is oriented north and so never gets such sunlight, cluttered house had moisture damage under the clutter... ), or that the advertisement standards do not allow generated image. I have no knowledge here though, but if I can come up with such potential legal showstoppers, prospective realtor can too.
@Mira I did research on this a couple years ago as a potential customer and it seems like a pretty crowded space to enter the market on (no single dominant company).
Also, from looking at demo work products the various companies that do the 3d work don’t work at the level of precision as Dreamworks. They seem to be high volume and are just placing objects and shadows artistically rather than geometrically/mathematically.
I wouldn’t want to spend time on this myself.