Lightroom preset ads are before/after first-last

Two stills of the same photo. Interpolate or a simple cut. A Lightroom preset ad is not a cinematic grade-from-nothing. It is the identical photograph twice: SOOC (or a flat export) as frame A, the preset applied as frame B. The product is the difference. Everything else in the frame should stay still.

That is a first-and-last-frame job, or it is a hard cut. It is not a text-to-video "moody photographer in a studio."

Same photograph, two grades

The honest pair:

1. Before. The photo the buyer already has — their RAW export, their JPEG, a demo image you are allowed to use. Do not generate a new "before" that is a different room, a different face, a different crop.
2. After. The same file with the preset on it. Apply it in Lightroom (or your editor). Export. That file is the product demonstration.

If you need the after still cleaned — a stray object, a colour you do not want in the demo — AI photo editor is instruction editing on the photograph you already have: describe the change, keep the photo. It is not a generator. Do not rebuild the scene.

Both stills must share crop, lens, subject, and resolution. First-last frame generation takes those two stills and builds the motion between them. Keep the frames in the same location and lens or the interpolate reads as a morph. Lighting is allowed to change — that is the preset — but the photograph underneath is not.

A pair generated as two unrelated "cinematic portraits" is not a preset ad. It is a morph between two people.

Interpolate, or just cut

Two legal finishes:

Interpolate. Create a first-and-last-frame transition video is the named job: attach start and end photos. The agent picks a first-last-frame-capable model. Output opens on your before and lands on your after. VEOs and Klings that take both frames, and Flux 3 first-last-frame, are that family. A paragraph with no stills will not run this job.

Write the prompt as the travel: a two-second grade settle, a slow exposure of the colour, not "a photographer walks through a loft." The stills already cast the shot. Recasting in text is how the face drifts off the demo file.

A simple cut. Before, then after. No generate. If the preset's sell is instant — the buyer needs to see the two states, not a journey — a cut is the honest ad. Do not buy an interpolate because motion feels more "content." Motion that warps the subject's jaw is a worse demo than a cut.

Loops (same image in both slots) are a different trick. A preset ad needs two different grades of the same photo. Using the after as both first and last gives you a living grade, not a before/after.

What wastes the demo

Generating the after. If the model invents the grade, you did not demonstrate the preset. You demonstrated a video model. Apply the preset. Then interpolate.

A talking creator explaining the sliders. That is a tutorial. This page is the ad: two stills, a join or a cut.

Mismatched crops. A 4:5 before into a 9:16 after is a different crop, not a grade. Match the files, then pick duration.

A last frame that is a different photo. "Same vibe, different wedding" is a morph. The buyer will not trust the preset. The glossary pitfall is choosing an end frame compositionally unrelated to the start. For presets, unrelated means a different photograph.

Approve both stills before you open the agent. If either file is wrong, every second of interpolate is waste. Edit the still. Do not reroll the join hoping the skin becomes the right skin.

FAQ

Do I have to interpolate? Can I just hard-cut?

You can just cut. Two stills on a timeline is a finished preset ad. Interpolate when the sell is watching the grade land. It is not required.

Can I prompt a video model to "apply my Lightroom preset"?

No. A generate does not load your .xmp. The after still is the preset, applied by you, then used as last frame. The model interpolates pixels. It does not run Lightroom.

Why not image-to-video from the before only?

Single-still I2V improvises the rest of the clip. It will not land on your approved after. First-last is how you force the landing. That is the whole point of two frames.

What if the after still needs a retouch?

Retouch the still on AI photo editor — one instruction, same photograph. Then run first-last. Do not ask the interpolate to remove a trash can that was in both frames.