Product scale across first last

Measure the pack in both stills. First and last frames need matching product scale or the morph looks like a grow ray. Viewers forgive mild motion. They do not forgive a bottle that inflates while a hand pretends nothing happened.

First-last workflows are powerful for product spots because they constrain endpoints. They punish sloppy endpoint design just as hard as they reward careful continuity.

Why scale drift reads as fake

Human perception is tuned for object constancy. When a labeled pack changes size relative to a hand or a table edge between endpoints, the interpolation path must invent a morph. That morph looks like a visual effect from an older music video, not like a camera move. Shoppers feel "ad trick" even if they cannot say "scale mismatch."

Hands make it worse. A hand that grips a small pack in frame A and a large pack in frame B forces finger topology chaos. Prop handedness errors compound the problem. Keep left and right consistent using habits from prop handedness left or right.

For Flux-oriented product spots, study Flux 3 first last frame product spots and the model entry for Flux 3 first last frame to video. Then build stills like a meticulous art director, not like a prompt gambler.

A measurement ritual before you generate

1. Place the product in both stills with the same relative size to a shared reference (hand width, table grid, torso).
2. Overlay the stills at 50% opacity in any editor and check label height.
3. Match camera distance cues: if frame A is a medium close-up, do not make frame B an unexpected macro unless the move is intentional and supported.
4. Match label orientation and fold shadows where possible.
5. Only then run the first-last generate.

If overlay shows a 10-20% size jump, fix the stills. Do not hope the model will "split the difference" gracefully. It might, and it might look cursed.

Keep lighting direction related across endpoints. Scale matched but lighting flipped still creates a harsh morph. Continuity is a bundle: scale, light, grip, and brand mark readability.

Shot design that uses first-last well

Best uses: a small camera push while the pack stays stable, a controlled rotation with matched size, a scene change where the pack remains the constant hero, or a before/after surface cleanup where the product holds steady as context changes.

Risky uses: teleports in pack size, sudden swaps to a different SKU silhouette, and hands that enter from opposite sides between endpoints.

Design the end frame first when the payoff is a clean pack shot for the grid. Then build the start frame as a variation that can honestly travel there. Ending on the hero often beats starting on the hero and wandering away.

Credits and reroll prevention

Scale mistakes are expensive because they look almost right until the last second of the clip. Reviewers who only scrub the middle miss them. Always scrub endpoints first.

Spend credits on corrected stills rather than on ten morphs from bad stills. Stills are cheaper emotionally and often cheaper in credits than motion rerolls.

Versely spend is in credits; protect them with overlays. If you need a workspace for product continuity tests, sign up for Versely and keep first/last still pairs named with measured scale notes in the filename.

The feeling to aim for is product respect. The pack should feel physical. Physical things do not inflate without cause.

Team review language that prevents grow rays

Replace vague notes like "feels off" with "pack height rose about 15% relative to the knuckles." Specificity gets stills fixed. Vagueness gets another generate that fails the same way.

Create a two-minute training clip for new artists: good matched endpoints versus grow-ray endpoints. Show the overlay technique. People remember the visual.

For multi-SKU shoots, do not reuse a first frame from SKU A with a last frame from SKU B because the bottles "are similar." Silhouette differences force ugly morphs. Pair frames per SKU.

If you must change pack size as a story beat (magic grow), telegraph it with design: sparkle, intentional effect, or cut. Accidental grow is the enemy. Intentional metamorphosis can be a style. The audience needs the intention to be readable.

QA on a phone. Desktop scrubbing underestimates how obvious scale drift looks in a hand-held vertical feed.

Store winning first-last pairs as templates. Reuse camera geometry and replace only the label artwork when SKUs differ only by flavor. Templates protect scale discipline across a catalog shoot.

When time is tight, fix the last frame first because it often becomes the thumbnail candidate. Then walk the first frame toward it. Thumbnail-worthy endpoints keep marketing and continuity aligned instead of fighting.

Photography notes for endpoint stills

Shoot or generate endpoint stills with the same focal length feel. Mixing ultra-wide distortion in frame A with a flatter lens feel in frame B creates morph stress even when ruler scale matches. Continuity includes lens language.

Keep the product label's vertical center at a similar height in frame when the camera is not meant to tilt. Unmotivated vertical travel of the pack reads as sliding geometry.

Avoid cropping the pack differently across endpoints unless the crop change is the point of the shot. A pack that loses its cap in one frame and regains it in another forces the model to invent plastic growth.

If you use AI to create the stills themselves, run the overlay check before motion. First-last video cannot fix still pairs that already disagree. Many "model is bad at products" complaints are still-pair problems.

Art directors should sign off on endpoint pairs the same way they sign off on print ads. Motion is not a place to invent missing discipline. The generate should be the easy part because the stills were hard in the right way.

If a client insists on dramatic size change without visual motivation, show a quick mock of the grow-ray risk using a 50 percent overlay. Seeing the mismatch usually unlocks permission to fix the stills.

Keep a printed or digital ruler in the set photo when you capture real pack plates for reference. Even a scrap of tape marked in centimeters helps artists match heights later. Physical measurement sounds old-fashioned because it works.

FAQ

How close do scales need to match?

Close enough that an overlay does not show an obvious jump. If you can see a clear size change at a glance, fix it before generating.

Can intentional zoom-like scale change work?

Yes when both frames support the move and the handoffs look photographic. Accidental mismatch is the problem, not motivated camera craft.

Do labels need to match exactly?

Match size, rough perspective, and readability. Exact pixel identity is less critical than believable constancy.