How to show a homeowner their renovated house from a photo, using AI concepts and a short video

A homeowner hands you a phone photo of a tired 1970s ranch and asks what it could look like. Now an AI image tool can show new siding, a bigger porch and black windows in a minute, and a short video can make it feel real. That speed helps you win the job and creates a new risk: the homeowner falls in love with something the AI drew, not something you designed or priced. This guide covers the tools, the photo to start from, what AI gets wrong about real houses, how to label the result, and how to turn it into a short before-and-proposed video.

Key takeaways

• Start from a straight-on daylight photo of the whole facade; AI redesign tools keep more of the real house when the photo shows it clearly.
• General image tools such as Gemini edit an uploaded photo from a text request, while contractor tools such as Hover build a scaled 3D model from smartphone photos and use real manufacturer products.
• AI redesigns routinely change rooflines, window sizes and the footprint without being asked, so compare every image with the real house before a homeowner sees it.
• Label every image and video "AI concept, not a final design or quote"; the FTC judges ads by their overall impression, pictures included.
• A short video works best as a "proposed" film of concept views with one real "before" photo for context, followed by drawings, samples and a written scope.

How can you show a homeowner their renovated house from a photo?

Take a straight-on daylight photo of the house, upload it to an AI image tool such as Gemini or ChatGPT, or capture the home in a contractor tool such as Hover, and ask for one change at a time: siding color, window trim, a new porch. Check each result against the real house, keep the versions that match what you could actually build, and label them as AI concepts. To make it feel real, sequence the before photo and the proposed views into a short video, then move to drawings and a written scope.

That last step matters as much as the first. The image starts a conversation; it doesn't replace design, permits or pricing.

Take a "before" photo the AI can work from

The redesign can only be as faithful as the photo. A dark, angled shot from the curb with a car in the driveway gives the AI room to guess.

• Straight on, from across the street, with the whole house in frame and a little sky above the roof. Hold the phone level so the walls don't lean (phone photography for architecture covers the settings).
• Overcast or soft daylight. Hard shadows hide trim and siding lines.
• Clear the frame. Cars, trash bins and hoses get "redesigned" too.
• Shoot the same spots every time. Take the same angles for each side you'll redesign, and for interiors shoot each room from the doorway. The same positions later give you the matching "after" photo.
• Measure separately. No image tool measures your house. Keep tape-measure or laser numbers, or a measured model, alongside the photos.

Ask the homeowner before you use photos of their home in marketing, even if you took them. A quick written OK covers the post you'll want to make once the job is done.

Tools that redesign a house from a photo

Two kinds of tools do this, and they suit different moments in a sale.

Google's help page says that with Nano Banana 2 you can "upload an image and ask Gemini to make edits," upload several images to create a new one, and make "local edits," described as "quick and specific changes to parts of an image," while editing images "isn't available to users under 18" (Gemini Apps Help, as of October 2026). ChatGPT works similarly; the ChatGPT images-to-video guide explains how to keep several views consistent.

Hover takes a different route. It says it transforms "smartphone photos of homes into scaled 3D models that can be used to see new colors and materials on the home," with a "materials library" of "products from major manufacturers," and lets contractors "estimate costs based on your designs" (Hover, as of October 2026). For an exterior job that will be quoted on siding, roofing and trim, a measured model tied to real products is a better basis for a promise than a generated picture.

Many firms use both: a quick AI image at the first meeting to find a direction, then a measured design for anything that goes into a proposal.

What AI gets wrong about a real house

An image model is trying to make a beautiful picture, not a buildable one. It changes things it wasn't asked to change. Check every image for:

• Massing and rooflines. A gable becomes a hip, a ridge rises, an attached garage grows by a bay.
• Windows and doors. Openings get bigger, move, or multiply. Those are structural and code changes, not cosmetic ones.
• Footprint and site. Porches deepen past setback lines, driveways widen, a tree vanishes from the neighbor's lot.
• Materials. Siding profiles, stone and trim that no manufacturer makes in that color, or that would never pass the local historic or HOA review.
• Landscaping and light. Mature planting and golden-hour light do a lot of the persuading, and neither comes with the contract.

Fix or discard images that show changes you wouldn't propose. If a homeowner asks for a feature the AI invented, treat it as a new request to design, price and, where needed, get approved: presenting to an HOA review board and similar boards look at drawings, not AI pictures.

Label it as a concept, every time

Put the label where the image goes: on the image itself, in the email and in the proposal. "AI concept for discussion. Not a final design, specification or quote." For video, put it in the caption and on the last frame.

This is about expectations first and advertising law second. The FTC says "advertising must be truthful and non-deceptive," and that it "looks at the ad in context - words, phrases, and pictures - to determine what it conveys to consumers," including "what the ad does not say" (FTC Advertising FAQs, as of October 2026). If you post a redesign publicly to market your services, a picture that looks like a finished job you did is a claim. Label concepts as concepts and finished work as finished work; AI rendering disclosure for architects has wording you can adapt. This is general information, not legal advice.

Make a short before-and-proposed video

A video does two things a still can't: it moves the homeowner through the house as if they were arriving, and it's easy to share with a spouse or a lender. Keep it honest and short.

1. Pick the images. One or two real photos of the house as it is today, plus concept views of the proposed design: the street view, the entry, a side elevation, a detail of the new materials and, if relevant, an interior.
2. Keep one design. Don't mix three color schemes in one film; make a separate video per option.
3. Decide the order you need. For a strict before-then-after sequence with a wipe or split screen, use a video editor; before-and-after renovation videos covers matched angles and transitions.
4. Or make a "proposed" film. Arch2Video takes exactly six images from one project, existing photos and AI concept images mixed, and returns a vertical film of about 30 seconds with an original soundtrack, for USD 15 per video. Its AI director plans the scene order, so it may not put the "before" photo first, and people in your photos aren't animated; a small cast of fictional adults appears in at least one scene. That suits a film of the proposed house in its real street, rather than a strict before-and-after comparison.
5. Label and review. Watch it through before sending; AI motion is an interpretation, and it can drift a roofline just as the still image could.

Don't promise what the AI drew

The risk isn't the image. It's the moment a homeowner signs believing they're buying it.

Write the scope from your drawings and specs, never from the picture, and say in the proposal that concept images are illustrative. When the job is done, photograph it from the same spots as the "before" photo; the real comparison is your best marketing (design-build video marketing shows how firms use it).

Frequently asked questions

Can AI show what my house would look like with new siding?

Yes. Upload a clear, straight-on photo to an AI image tool and ask for one change, such as a siding color or profile. The result is a quick idea, not a product match: check it with real samples. Contractor tools that build a 3D model from your photos and use real manufacturer products give a more reliable preview.

Is it okay to send a homeowner an AI-generated redesign of their house?

Yes, if it's labeled as a concept and you've checked it for changes you wouldn't propose, such as bigger windows or a new roofline. Put "AI concept, not a final design or quote" on the image and in the email, and follow it with drawings and a written scope before anything is signed.

Can I post AI redesigns of clients' houses on social media?

Get the homeowner's permission first, since it's their home and often recognizable. Then label the post clearly as an AI concept so it can't be mistaken for completed work; the FTC looks at the whole impression of an ad, pictures included. Never present an AI image as a finished project you built.

How do I make a video from a photo of a house and an AI redesign?

For a strict before-and-after with a wipe or split screen, line up matching photos and concept images in a video editor. For a short film of the proposed design, a six-image tool works: Arch2Video takes six images from one project, mixing real photos and concept images, and returns a 30-second vertical video for USD 15, with the scene order chosen by its AI director. Label it as a concept either way.