Before and after videos: what is safe to make with two photos

A before and after video is one of the clearest kinds of proof a tradesperson can post. The customer sees the stained floor, then the clean floor. The problem with AI tools is that they can create the "in between" so smoothly that the video looks like a real transformation. If a viewer can read the middle frames as evidence of how the work was done, that is a misleading picture, and it can look like a false claim.

The answer to the question "how do I make one from two photos" is mostly simple and free. This guide gives that answer first, then covers shooting matched photos, consent, captions, the few places AI helps, and where Higgsfield, a platform with several AI image and video models behind one login, fits. It is for plumbers, cleaners, painters, landscapers, repair shops and the agencies that serve them. It is general information, not legal advice.

This page may contain affiliate links from which Blur Studio earns commissions.

The free method: a wipe between two real photos

Most trades do not need AI for this. Here is the whole method:

1. Take the before photo and the after photo from the same spot.
2. Open a free editor. Canva, a design tool that also makes simple videos, has a Free plan, and the basic wipe and slide transitions do not need a paid plan. CapCut, a video editor from the makers of TikTok, is a common alternative; check each pricing page for what the free version includes.
3. Put the before photo first, then add a wipe, slide or soft fade into the after photo.
4. Add the words "Before" and "After" as text in the editor.
5. Export.

A simple shape from Runway's own guide is about 2 seconds of the before, 2 seconds of the transition, and 3 seconds of the after. That is only a guide. Keep it short enough that the reveal lands.

That is the whole job. Both photos stay untouched, it costs nothing, and it passes the honesty test.

Why a real reveal beats a generated one

Runway, an AI video tool that sells an AI transition, says in its own guide that an AI camera move on each real photo often reads as more believable than an AI morph. Cliprise, another AI video vendor, gives a rule I like: if the claim cannot appear on your landing page or sales page, do not put it in the video. The US Federal Trade Commission says an ad's pictures are read as claims, claims need evidence, and the impression a reasonable consumer takes is what counts. I found no FTC page that deals with AI before and after images, so I rely on that general truth-in-advertising guidance. Check your state and local consumer rules as well.

My hard line is simple. A generated morph shows a transformation that never happened frame by frame. If the claim is the result of your work (a stain removed, a fence repaired, a lawn rebuilt), do not let a model invent any part of it.

Four transition styles and when each fits

Runway's guide and Cliprise's workflow page, both from AI video vendors, describe similar styles. Here is how I would match them to trades work.

Style	What the viewer sees	Best for
Wipe or slide	A line moves across and reveals the after photo	One strong reveal where both photos share the same framing. My default for trades.
Split-screen	Both photos side by side at once	Small changes the viewer needs to compare directly, such as grout or paint edges.
Hard cut with labels	Before, then after, with "Before" and "After" typed in the editor	The most plainly honest option. Cliprise calls it clear and honest.
Dissolve or AI morph	One photo blends into the other	Soft cosmetic upgrades at most. An AI morph invents the middle, so I keep it away from proof of work.

How to shoot matched photos on the job

The wipe only looks good if the two photos line up. Use this shot list on site:

1. Pick the spot before you start work. Mark it with a piece of tape on the floor or wall, or take a reference photo on your phone to match later.
2. Use the same camera position, angle, distance and zoom both times.
3. Shoot at the same time of day if you can, so light is similar.
4. Wipe the lens.
5. Hold the phone upright for social posts.
6. Add one wide shot for context, showing the whole room or yard.

Runway's guide gives the same advice: match camera position, angle, distance and lighting. It is the cheapest upgrade you can make, and it matters more than any tool.

Customer consent and privacy

This is the part most before and after guides skip. Your photos often show a customer's home interior, house number, family photos, children, cars or computer screens. Before you post:

• Get the customer's written permission to publish the photos. A short line in your quote or invoice is enough to start. Ask a lawyer about wording for your area.
• Crop or blur house numbers, street signs, plates, documents and anything with a face.
• Do not tag the street address. Use the suburb or area.

If you upload a customer's photos to an AI tool, you are also sharing them with a third party. Higgsfield's terms say it may use content to train and improve its models, except under Enterprise agreements. That is one more reason to keep customer photos out of AI tools unless the customer has agreed.

Captions, labels, dates and keeping originals

Add labels in the editor, not through the AI model, because models can mangle text. A caption template:

• Job type: "Bathroom regrout".
• Area: "Northside" (not the street).
• Date: "March 2026".
• "Photos unedited", only if it is true.

If you did enhance anything, even light fixes, say so honestly. Keep the original files with their date and a job number, along with the customer's consent. If anyone ever questions the video, you can show the originals.

Also avoid before and after posts for health, beauty or weight claims. Those are a risky category, and the FTC is strict about health claims.

Options for adding motion

1. A push-in or pan in the editor. Add a slow zoom on each real photo. Free and honest.
2. A phone video of the finished site. This is better proof than any two-photo reveal. Walk slowly through the clean room for ten seconds.
3. A slider on your website. Many website builders have a before and after slider that lets the visitor drag. It is good for a single project page.
4. AI motion on each real photo. If you want a gentle camera move, an AI tool such as Higgsfield's image-to-video tools can add one to each photo separately. Runway, an AI video tool with its own models, lists Free at 125 one-time credits, Standard at $15 a month ($12 billed annually) for 625 credits, and Pro at $35 ($28 annual) for 2,250 credits, with a 5 second Gen-4.5 clip costing 60 credits (checked 2026-11-03).
5. An AI transition between the two photos. This is the option I would not use as proof.

For when a basic editor is enough, see Higgsfield vs Canva and CapCut. For the real-versus-generated question in general, see AI video vs stock footage vs filming yourself.

Where Higgsfield fits, and where it does not

Higgsfield lets you upload a still picture and animate it into a short clip with the video model you choose. Its help center describes DoP (short for Director of Photography, Higgsfield's own image-to-video model), which turns one image into a 3 or 5 second clip with preset camera moves. In its plan comparison table, control of the start and end frame, which lets you pick a first and last picture, is listed on Plus and Ultra only. Check the pricing page for the current gating. Plans are Starter $19 a month for 270 credits, Plus $59 ($47 annual) for 1,200, and Ultra $129 ($99 annual) for 3,000. A credit is the unit the platform charges for each generation. The pricing table lists DoP at roughly 3 to 7 credits for a 3 second clip, and the exact cost shows on the Generate button.

In theory you could use the before as the first frame and the after as the last frame and let the model make the middle. That is exactly the part I would not use as proof. So my advice for trades is narrow:

• Use Higgsfield, if at all, for a gentle camera move on each real photo, or for the intro and outro of a longer video, not for the reveal.
• Free outputs are watermarked and the Free plan has 0 credits. Commercial use is marked "Not included" on Free in the pricing table, while the help center says use is not restricted to paid plans. For a business page, use a paid plan.
• If you need nothing but the reveal, skip the paid tool entirely. The free editor is better.

See my Higgsfield review for the wider picture, and video transition effects if you want stylish cuts in other kinds of videos.

Safe and unsafe: what you may change

Action	Safe?
Wipe or fade between two untouched photos	Safe. Keep originals.
Crop to the same frame, adjust nothing else	Generally fine, say so if you edited.
Add "Before" and "After" text in the editor	Safe.
Slow zoom on each real photo	Safe.
AI camera move on each real photo	Acceptable if you check every frame and label it honestly.
AI-generated in-between or morph	Risky. It shows a change that did not happen.
Cleaning up or improving the after photo with AI	Do not. It is a false claim about your work.

For the editing tools themselves, and why they are risky on proof photos, read Higgsfield upscale, relight and edit tools.

Cost if you do generate

The editor method has no generation cost. If you add an AI transition anyway, here is the arithmetic with made-up numbers. At Runway's published 60 credits for a 5 second Gen-4.5 clip, one generated transition per job is 60 credits. With 3 tries that is 180, so Standard's 625 credits cover about three jobs in a month. For most trades that cost buys little. On Higgsfield, using the pricing-table figure of about 7 credits for a 3 second DoP Standard clip, 3 tries is 21 credits, but the real cost shows on the button. See how to spend Higgsfield credits wisely and Higgsfield pricing.

A 14-day test plan with pass and fail

This is a plan you can run. Small accounts give directional data only.

1. Pick two similar jobs.
2. Post the plain editor wipe for one job. For the other, post a version with AI camera motion on each photo (not a generated morph) if you want to test AI at all.
3. Post both on the same channel, such as your Google Business Profile or Facebook page.
4. After 14 days, compare enquiries (calls and messages) that came from each post.
5. Pass: no comments calling it fake, and enquiries from the AI-motion post are not lower. Fail: any comment questioning the result, or fewer enquiries. If anyone mentions AI in a bad way, stop. Go back to the editor wipe.

Who should skip AI transitions entirely

• Anyone whose before and after is the proof of a result: stains, repairs, rebuilt lawns, paint jobs.
• Health, beauty and weight businesses, where claims are tightly controlled.
• Anyone without written customer permission to publish the photos.
• Anyone who plans to touch up the after photo.

How I'd approach this

I would put a short shot list in every crew's truck: tape mark, same angle, wipe the lens, wide shot, written consent. Then I would make every reveal with a free editor and post the originals' dates in the caption. If I wanted to experiment with AI, I would restrict it to camera moves on each real photo and test it against the plain version for two weeks. And if AI ever made the work look better than it was, I would delete the clip.

If you run a crew or an agency and want the photo capture, consent, approval and posting steps in one repeatable flow, Blur Studio builds approval portals and shared asset libraries around these tools; see the business page. Most trades will do fine with a shot list and a shared folder.