AI Before and After Ad Video Generator

A before-and-after is a claim that one thing changed. The moment there is a cut, the viewer files it as two things filmed separately — which is why this format has always been shot as a single unbroken take, and why it has always been the expensive one.
Fifteen continuous seconds, sound included, from a paragraph. Your first one is free. No card.

Write the change

0 / 7000

This run spends credits — the free clip is 4s · 768P · 16:9.

Sign up free and your first MiniMax H3 clip renders 4s · 768P · 16:9, with sound. A plan raises the ceiling to 15s, 2K and four at a time — the free clip is a size, not a different model.

  • Name what must not move. A change is only readable against something held still — a position, a stem, a badge, a horizon
  • Find the frame where both states already look alike and describe that resemblance. That frame is the transition; everything else is decoration
  • Write the prohibition out: no hard cut, no black frame, no splice, no obvious effect. The model will take a cut if you leave it the option
  • Give the end card one line in quotation marks. A list of tagline options gets you one of them, rendered badly

No card: every ad on this page plays without an account, and a free account runs one of your own — 4s, 768P, with sound, no watermark, yours to download and to use commercially.

Five transformations and the complete prompt behind each. Every card opens the whole thing.

Before and after in one shot: the cocoa on a coffee’s foam becomes the ridge of a sand dune without a cut

MiniMax wrote this one

The brief on the left. The take on the right.

Five transformations with the whole prompt printed, not a summary

MiniMax wrote this one16:9 · 15s · 2 reference images · official MiniMax prompt

MiniMax’s own prompt, and the reason this page exists. It does two things nothing else here does. First it names the resemblance — the cocoa particles, the foam undulations and the liquid swirls are described as being like the dune ridges and wind erosion in the second image — so the model is told where the join is rather than left to invent one. Then it closes every exit: Strictly avoid screen cracking, no black screen, no hard cuts, no obvious special effects, no sense of splicing, and finally One continuous shot, no splicing in the middle. Its audio comes back at 32 kHz, which is this model’s own fingerprint.

@Image 1 The shot quickly pushes into the detail of the coffee surface's milk foam, cocoa powder particles, and dark coffee liquid texture, letting the surface particles, foam, and liquid ripples gradually cover the entire screen, obscuring everything else. The shot maintains a realistic macro texture, extremely shallow depth of field, with fine powder slowly falling in backlight, the surface texture like a transition between fine sand and fluid. Avoid screen cracking. At the moment the coffee powder particles, milk foam undulations, liquid swirls are highly similar to the sand dune ridges, wind erosion textures, and blowing sand particles in @Image 2, naturally and seamlessly transform into a desert landscape. Then the shot continues to push forward, revealing the full view of the sand dunes from @Image 2. Strictly avoid screen cracking, no black screen, no hard cuts, no obvious special effects, no sense of splicing. Overall, it should be real, quiet, restrained, like the same granular material naturally changing from the coffee's microscopic surface into a macroscopic desert world. One continuous shot, no splicing in the middle.

Before and after in one shot: the cocoa on a coffee’s foam becomes the ridge of a sand dune without a cut
Stock to street weapon, on a clock16:9 · 15s · text only · 7 timed beats

The literal version of the format — FROM STOCK TO STREET WEAPON — laid out as seven blocks of two seconds each, from the dusty garage to the wet-asphalt hero shot. Worth noting what it gets away with: it names a real make and model, it opens by asking for 9:16 and comes back wide, and its closing card, SWIFT — REBORN, arrives clean. One line of end text, given once, in caps, with nothing to choose between.

FORMAT: 15-second cinematic vertical car transformation film 9:16 aspect ratio Ultra-realistic automotive commercial aesthetic Japanese street culture + luxury cinematic realism High-energy transformation pacing CONCEPT: “FROM STOCK TO STREET WEAPON” A normal Maruti Suzuki Swift slowly transforms into a fully modified street machine through cinematic garage sequences, mechanical close-ups, aggressive lighting, and dramatic reveal shots. TIMELINE: 0:00–0:02 Old stock white Maruti Suzuki Swift parked inside a dim industrial garage. Dust particles in sunlight. Static realistic atmosphere. Soft engine ticking sounds. 0:02–0:04 Quick cinematic close-ups: tools dropping on metal table, garage lights turning on, mechanic gloves tightening, car jack lifting vehicle. Fast-paced editing begins. 0:04–0:06 Transformation montage: new alloy wheels installed, wide tires mounted, lowered suspension, custom front bumper attached, black spoiler fitted. Mechanical sparks and cinematic motion blur. 0:06–0:08 Paint transformation sequence. Original white paint peels away into glossy metallic midnight black. Water droplets and reflections across body panels. Neon garage lighting reflects on paint. 0:08–0:10 Interior upgrade shots: digital dashboard glowing, sport steering wheel, bucket seats, ambient red lighting, push-start ignition. Engine rev echoes. 0:10–0:12 Aggressive cinematic reveal: modified Swift exits garage slowly. Headlights turn on through smoke. Low-angle tracking shots. Wet asphalt reflections. Turbo flutter sound. 0:12–0:15 Final hero sequence: car speeding through empty city streets at night, rolling shots, drifting corner, rain reflections, motion blur, camera orbit around vehicle. End with parked hero shot under neon lights. FINAL TEXT: “SWIFT — REBORN.” VISUAL STYLE: Fast & Furious-inspired realism Moody cinematic lighting Japanese underground street-racing atmosphere Wet asphalt reflections Practical garage lighting High-detail metallic textures Subtle film grain Deep contrast shadows Realistic smoke and sparks CAMERA: ARRI Alexa Mini LF aesthetic Anamorphic lens look Low-angle tracking shots Wheel close-ups Slow-motion mechanical shots Drone-style orbit reveal Handheld garage realism SOUND DESIGN: Metal tool impacts Hydraulic jack sounds Engine revs Turbo flutter Deep bass transitions Tire screeches Cinematic trailer hits Aggressive phonk-inspired background energy QUALITY: Masterpiece-level automotive realism Luxury commercial production quality Pinterest-saveable cinematic car edit Ultra-detailed reflections and textures Feels like a premium car modification advertisement

Before and after ad: a stock hatchback in a dim garage becomes a lowered black street car on wet neon asphalt
Problem, then product, no clock at all16:9 · 15s · text only · no timestamps

No timestamps anywhere. One paragraph of prose — a man wilting in the heat, a sip, and the road turning green around him — and the transformation still lands, because the *subject* never moves. What does not survive is the tagline. The prompt ends with four options under the heading Tagline ideas:, and the model rendered one of them with a word missing: the clip says Beat the Heat. the Freshness. Give it a menu and it will order badly.

A young person walks under the blazing summer sun, looking exhausted and sweating heavily. The road shimmers with heat waves, and everything appears dry and dull. Suddenly, they grab a chilled bottle of premium fruit juice from a cooler and take a refreshing sip. Instantly, the environment transforms—lush green trees bloom, vibrant flowers appear, a cool breeze flows, water splashes through the air, and glowing particles surround the scene. Ice cubes and fresh fruit slices (orange, mango, or according to the flavour) swirl around the bottle in cinematic slow motion. End with a stunning close-up of the juice bottle covered in cold water droplets against a bright, refreshing background. Ultra-realistic, premium commercial, 4K, cinematic lighting, high-detail, smooth camera movements, vibrant colours, luxury beverage advertisement. Tagline ideas: Beat the Heat. Taste the Freshness. Every Sip Brings Life. Refresh Your Day, Naturally. Stay Cool. Stay Fresh.

Before and after ad: a man wilting in summer heat, then the same road green and cool around a chilled juice bottle
Three acts and a tagline that survived16:9 · 15s · text only · 3 acts

The control case for the sentence above. Three acts of five seconds, and exactly one piece of end text, written once and in quotation marks: the NOVA X logo appears elegantly with the tagline: “NOVA X — The Future in Your Hands.” It comes back set correctly and spelled correctly. The difference between this end card and the juice one is not the model. It is that this prompt made the decision instead of delegating it.

Create a 15-second premium cinematic branding TVC for a futuristic smartphone brand called NOVA X. 0–5 seconds: Extreme close-up of a sleek futuristic smartphone resting on a glossy black surface. Dramatic studio lighting sweeps across the device, revealing its ultra-thin design and premium metallic edges. Cinematic camera push-in, macro product details, realistic reflections. 5–10 seconds: The phone activates with a stunning futuristic interface. A young professional picks it up, and the environment transforms into a vibrant digital world with floating holographic visuals, smooth UI animations, and seamless interaction. Dynamic camera movement, premium commercial aesthetic. 10–15 seconds: Hero product shot of the smartphone standing vertically against a minimalist futuristic background. The NOVA X logo appears elegantly with the tagline: “NOVA X — The Future in Your Hands.” Cinematic lighting, high-end TV commercial quality, photorealistic, ultra-detailed, smooth motion, polished advertising style, 4K.

Before and after ad: a phone lying flat under a moving key light rises to a lit hero shot with its tagline set beside it
A shot list that is not valid JSON16:9 · 7s · text only · a JSON shot list

The shortest ad here at seven seconds, and the prompt is a JSON shot list — camera angle, subject, action, visual details and audio cues as fields, three shots deep. It is also not valid JSON: there is a trailing comma before the closing bracket, so nothing could have parsed it. That is the point. H3 reads one text field, so the structure is doing its work on the language model rather than on a parser, and a schema you invented yourself is worth more than prose even when it is broken.

{ "shots": [ { "shot_id": "1", "start_time": "00:00", "end_time": "00:02", "camera_angle": "Close-up", "subject": "Strawberry", "action": "A single, dull strawberry transforms into a vibrant, ripe strawberry, floating against a pink background.", "visual_details": "Initial strawberry is muted, then becomes intensely red with visible seeds; pink background; subtle glow effect.", "audio_cues": "Light, subtle, almost magical sound effect." }, { "shot_id": "2", "start_time": "00:02", "end_time": "00:05", "camera_angle": "Medium shot", "subject": "Strawberry and beverage can", "action": "The vibrant strawberry transforms into a white and pink beverage can with 'STRAWBERRY CREAM' text. Liquid pours into the can from above, and strawberries/water droplets float around it.", "visual_details": "Can is sleek, covered in condensation; red liquid splashing into the can; numerous floating strawberries (whole and sliced) and water droplets; pink background.", "audio_cues": "Sound of liquid pouring, bubbling, and light splashes; upbeat, commercial-style music." }, { "shot_id": "3", "start_time": "00:06", "end_time": "00:10", "camera_angle": "Close-up", "subject": "Top of the beverage can", "action": "Focus on the top of the can as the liquid continues to pour and splash, with droplets and strawberries in motion.", "visual_details": "Intricate details of the can's opening and condensation; dynamic splashes of liquid; close-up of floating strawberry pieces.", "audio_cues": "Enhanced sounds of liquid splashing and bubbling; continued music." }, ] }

Before and after ad: a dull strawberry ripens and becomes a condensation-covered drink can in seven seconds

What this is, and what it is not

The cut is where the claim dies. So there is no cut.

This is not a slideshow with a wipe between two photographs. Split-screen and swipe transitions read as *two assets*, which is exactly the reading a before-and-after cannot afford — and it is why the good ones have always been a locked-off camera, a real change, and one unbroken take.
What arrives here is fifteen seconds of continuous frame with its own sound. That is the whole ad, not a shot from it.

The reason this format costs what it does is that the unbroken take is either a real shoot you have to stage or a morph somebody has to build. A social ad through an AI studio is quoted at $1,000–$3,500 against $6,000–$20,000 shot traditionally, and a single 30-second spot runs $15,000–$30,000 the old way. At shot level, VFX starts around $3,000 for something as small as a wire removal.
The part nobody budgets is the second attempt. A studio revision round is $2,000–$5,000 traditionally and $200–$800 at an AI shop, and vendors will tell you a shot through five rounds can end up costing what a hard shot costs delivered first pass. Here a re-run is 125 credits, about $2.75, and a failed generation is refunded. An extra aspect ratio, quoted at $500–$1,500 elsewhere, is the same per-second rate.

A 30-second spot, shot traditionally
$15k–$30k

A 30-second spot, shot traditionally

One revision round at a studio
$2k–$5k

One revision round at a studio

An extra aspect-ratio cut, elsewhere
$500–$1,500

An extra aspect-ratio cut, elsewhere

A fifteen-second take here, sound included
~$2.75

A fifteen-second take here, sound included

  • No morph to build by hand
  • No stock music licence to buy
  • No watermark on any tier, including the free one
  • No surcharge for the second ratio, or the third

What you put in

Four ways to hand over the two states

The question this band answers is how much of the before and the after you already own. The more of it you attach, the less the model has to guess — and in this format a guess is the difference between a transformation and two unrelated shots.

  • Give it both ends

    The before in one image, the after in another, and a sentence about the frame where they resemble each other. This is how MiniMax’s own example works and it is the most controllable route on the page.

    Both states pinned · up to 9 images
  • Give it the before only

    Your real product, room or packshot as the opening frame, and the after written out. The start is exact and the finish is described, which suits anything where the *before* is the thing a customer recognises.

    Start from your own frame
  • Write both ends

    No upload. Fastest way to find out whether a transformation reads at all before you commit real assets to it — four of the five ads above went this way, including the one with seven timed beats.

    Concept first · text to video
  • Hand it a transition you like

    A clip off your swipe file whose morph works, plus your subject. It takes the camera behaviour and the timing of the change from the clip rather than the footage.

    Borrow the move · up to 3 clips

All four bill the same: 8 credits a 768P second, 13 at 2K, plus 5 for the clip. The first five reference images are free and the sixth onward is 4 credits each. Every limit and rate, with the date each last changed.

Start to finish

How to make a before-and-after ad in four steps

The first two are the whole craft and they cost nothing to get right. Both are decisions about the *middle* of the clip, which is the part every brief forgets to specify and the only part the audience is actually judging.

  1. A single strawberry held dead centre against flat pink, the anchor the rest of the clip changes around

    Decide what is not allowed to move

    A change is only legible against something held still. In the ads above it is a strawberry that keeps exactly the same screen position while everything around it changes, a single stem that stays anchored, a badge that stays on the grille. Name that thing first, in one clause, and the transformation has something to be measured against.

    One anchorNamed in one clause
  2. Cocoa powder across coffee foam at macro, one frame before it becomes a desert

    Find the frame where both states already look alike

    This is the transition, and it is a matching problem rather than an effects problem. Cocoa particles on foam look like wind erosion on a dune; a wet garage floor looks like wet asphalt. Describe the resemblance and the model has somewhere to put the join — leave it out and it will reach for a dissolve.

    Match the textureNot a dissolve
  3. The finished car rolling out of the garage through its own smoke, still in the same take

    Strictly avoid screen cracking, no black screen, no hard cuts, no sense of splicing… one continuous shot, no splicing in the middle.

    PictureSound

    Forbid the cut, in writing

    Say it as a prohibition and say it more than once: no hard cuts, no black screen, no sense of splicing, and then one continuous shot. Sound is generated with the picture, so name the transient the change makes as well — the transformation wants a noise, and a silent morph reads as an edit even when it is not one.

    Say it twiceOne field, both channels
  4. The finished phone hero shot with its tagline set beside it, the end of the take
    Audio track · 32 kHzAFTER-01.MP4

    Test at 768P, master the take that lands at 2K

    A 768P second is 8 credits against 13 at 2K, and upgrading a finished 768P clip afterwards costs exactly what rendering it at 2K would have. In a format where the transition either works or does not, the attempts you throw away are most of the job — and a failed generation is refunded automatically.

    768P · 8 credits a second2K · 13 credits a second

Four steps, no install, no card. Open the generator.

The thing nobody tells you

The five that cost a render. Four of them are visible on this page.

Every item below came out of the five ads above rather than off a spec sheet, and four of them you can check yourself by playing the clip next to the prompt that made it.

  1. 01A menu of taglines gets you a bad one

    The juice prompt ends with four options under Tagline ideas:. The model picked one and dropped a word from it — the finished frame reads “Beat the Heat. the Freshness.” The phone ad gave exactly one line, in quotes, and it came back correct. Decide the words before you generate them.

    one line, in quotes

  2. 02There is no cut to hide behind

    Fifteen whole seconds at 24 fps is the ceiling and there is no stitching in the product, so the transformation has to be one take. That is the format’s natural shape rather than a limitation — but it does mean an ad with two changes in it is two generations, and they will not match.

    4–15s, one take

  3. 03A structure beats prose, even a broken one

    The seven-second strawberry ad was written as a JSON shot list with a trailing comma in it — invalid, unparseable, and it worked anyway. H3 takes one text field, so a schema you made up is doing its work on the reader rather than on a parser. Use tables, use fields, use timestamps.

    fields, not paragraphs

  4. 04The prompt cannot set the shape

    The car ad opens by specifying a 9:16 vertical film and came back 16:9, because the ratio is a control on the generator and the sentence is just more description. Set it on the chip above. Every shape bills at the same per-second rate.

    ratio is a chip

  5. 05Land the cut on 8.000 seconds

    H3 renders on a 17n+5 frame grid at 24 fps, so most lengths come back a fraction long — 15s is really 15.083s. 192 frames, exactly 8.000 seconds, is the only whole second in the range, which makes it the length that cuts to a music bed without drifting.

    192 frames = 8.000s

A fifteen-second take at 768P is 125 credits, about $2.75. A revision round at a production studio is $2,000–$5,000, which is the comparison that matters in a format where you are buying attempts.

Why this and not a transition pack

A morph is a sound as much as it is a picture

Watch any good transformation with the audio off and the join stops working. The change needs a transient on the frame it happens — a swell, a whoosh, a hard mechanical hit — and on a template tool that is a second purchase, dropped in afterwards and nudged into place by hand.

  • One pass, both channels

    The transition arrives with its own noise

    32 kHz stereo generated alongside the picture, so the transient lands on the frame the change lands. Name the swell, the hit and the room; there is no library to license and no sync pass.

  • No stitching

    Fifteen seconds of unbroken frame

    Not four clips and three transitions — one continuous take at 24 fps, which is the only structure this format actually believes. It is also why the ceiling is not really a ceiling: a before-and-after that runs past fifteen seconds has usually stopped being one.

  • Twelve files in

    Both ends, pinned

    Up to 9 reference images, 3 clips and 3 audio files, 12 total. That is the before, the after, the colourway and a swipe-file clip for the timing, in one request. Most hosted doors onto this model do not publish their cap.

  • No seat

    You pay for attempts, not for a month

    No monthly floor and nothing expires for a year. This is 8 credits a second with a refund on anything that fails, which in a format where the middle either works or does not is the number that decides how many times you get to try.

What you give up

A real result, and the proof of it. This makes a picture of a change; it does not make the change, and it cannot tell you the product caused one. If the ad turns on an outcome a customer experienced, that outcome has to be real and documented before anything is generated — the band below is about exactly that. Frame-exact control is the other thing you give up: you can specify the join but you cannot art-direct it to the frame.

What it actually costs

A 15-second take at 768P is 125 credits. About $2.75.

A clip is a flat 5 credits for the prompt-understanding pass, plus 8 credits per 768P second or 13 per 2K second. The first five reference images are free, so pinning both ends of the transformation adds nothing. One credit is one cent of MiniMax’s own published list price, so the table can be checked against a page this site does not control. A generation that fails is refunded automatically — which is the line that matters here, because in this format you are buying attempts at a join.

What a transformation costs, by length

One 4s morph, 768P
37 credits
One 8s ad, 768P
69 credits
One 15s ad, 768P
125 credits
One 15s ad, 2K
200 credits
Six 15s attempts, 768P
750 credits
The keeper again at 2K
200 credits

Full pricing · every limit and rate, dated · what a 2K second really costs

Same monthly credits either way

  • FreeNo card

    One real MiniMax H3 clip, with sound. Everything past 4s · 768P runs on credits.

    $0forever

    No credit card at any point

    Start free

    Bot check only — no email

    1 clip on MiniMax H3

    MiniMax H3 · 4s · 768P · 16:9 · with sound
    Check in 7 days — 37 credits, one more clip

    Any pack from $9.9 unlocks the MiniMax H3 AI video generator — every mode, 768P and 2K

    • MiniMax H3 at native 2KPaid only
    • Native audio with the picturePaid only
    • One clip, no card
    • 1 clip at a time
    • Failed clips cost nothing
    • Private generation
    • Any pack unlocks the MiniMax H3 AI video generator on every mode

    Free is a different engine. See plans

    Speed & queue

    Queue
    Free lane
    Jobs at once
    1
    Batch
    1
    History kept
    24h · 7 days signed in
    Support
    Community
  • LiteSave 30%

    Unlock the MiniMax H3 AI video generator at native 2K. The smallest paid plan — Pro is the one most people pick.

    $16.9/mo$24.9

    $202.8 billed yearly · Save $96 a year

    7-day refund · cancel anytime

    750 credits / month · ~20 videos

    4s · 768P · text to video

    or ~13 at 4s 2K

    Same credits monthly or yearly

    • MiniMax H3 at native 2K — with native audio
    • Dialogue, effects and music in the same file
    • Up to 15 seconds15s
    • ~20 videos a month at 4s 768P, text to video
    • Failed clips cost nothing
    • 7-day refund if credits are unused
    • 1 job at a time
    • Batch 2 variants of one prompt

    Commercial use included — what that covers

    Speed & queue

    Queue
    Standard
    Jobs at once
    1
    Batch
    2
    History kept
    7 days
    Support
    Email · 48h
  • Recommended for most
    ProSave 30%

    $32 more than Lite. ~47 videos a month, same MiniMax H3 AI video generator, faster queue.

    $39.9/mo$56.9

    $478.8 billed yearly · Save $204 a year

    7-day refund · cancel anytime

    1,775 credits / month · ~47 videos

    4s · 768P · text to video

    or ~31 at 4s 2K

    Same credits monthly or yearly

    • Everything in Lite
    • ~47 videos a month at 4s 768P, text to video
    • 3 jobs at once
    • Video-to-prompt
    • Batch 4 variants of one prompt
    • History kept for 7 days
    • 7-day refund if credits are unused

    Commercial use included — what that covers

    Speed & queue

    Queue
    Fast
    Jobs at once
    3
    Batch
    4
    History kept
    7 days
    Support
    Email · 12h
  • StudioSave 30%

    A full month of volume, first in line, and a human who answers in 4 hours.

    $99.9/mo$142.9

    $1,198.8 billed yearly · Save $516 a year

    7-day refund · cancel anytime

    4,425 credits / month · ~119 videos

    4s · 768P · text to video

    or ~77 at 4s 2K

    Same credits monthly or yearly

    • Everything in Pro
    • ~119 videos a month at 4s 768P, text to video
    • 8 jobs at once
    • Batch 4 variants of one prompt
    • History kept for 7 days
    • Priority support in 4 hours
    • Lowest credit burn we offer
    • 7-day refund if credits are unused

    Commercial use included — what that covers

    Speed & queue

    Queue
    Priority — first in line
    Jobs at once
    8
    Batch
    4
    History kept
    7 days
    Support
    Priority · 4h

The most regulated shape in advertising

A before-and-after is a claim, and a generated one is still a claim

No format attracts more scrutiny than this one, because it states a result in pictures without ever writing the sentence down. Generating it does not move it out of that category — it moves it further in, because now the “after” never happened to anybody.

  • Commercial useIncluded, on the free clip as well. No watermark on any tier, no per-clip royalty, no separate licence to buy. MiniMax claims no rights over the Outputs you generate and this site adds none of its own.
  • The result has to be realA depicted outcome is a performance claim, and the rules on substantiating one do not soften because a model drew the frame. If a typical customer will not get what the clip shows, the clip is the problem — and “typical” is doing the work in that sentence.
  • Bodies, skin and teethHealth, weight and cosmetic transformations are the sharpest end of this and several regulators treat them as their own category. Do not generate a body result. If you have real before-and-after photography with consent, that is a different job and a different tool.
  • Say that it is syntheticMeta, TikTok and YouTube all require realistic synthetic media to be disclosed and each has the control in its upload flow. A photoreal transformation is precisely what those rules were drafted against.

Responsible use in full

Before you spend anything

AI before and after ad video generator FAQ

The eight questions people ask before running a transformation ad, answered with the numbers behind them.

Is this before and after video generator free?

The first clip is, and it is the real model — 4 seconds, 768P, with sound, no watermark, no card, and it never expires. Every ad on this page plays without an account. After that an 8-second take is 69 credits and a 15-second one is 125; credits start at $9.9, which puts a full ad at about $2.75. A generation that fails costs nothing.

How do I stop it cutting instead of transforming?

Two sentences do most of the work. Describe the frame where the two states already look alike — MiniMax’s own example says the cocoa particles and foam ripples resemble dune ridges and wind erosion — and then forbid the alternatives explicitly: no hard cut, no black frame, no splice, no obvious effect, one continuous shot. Both sentences are printed in full in the first showcase card above.

Can I make a 30-second before and after?

Not as one take. The range is 4 to 15 whole seconds at 24 fps and there is no stitching in the product, so 30 seconds is two generations and an edit — and the join between them is exactly the cut this format cannot afford. In practice a transformation that needs 30 seconds is usually two transformations.

Can I use my own before and after photos?

Yes — attach both, and say which is which. Up to 9 reference images in one request, the first five free, so pinning both ends costs nothing extra. Bear in mind the model redraws every pixel: what comes back resembles your photographs closely rather than reproducing them, which is fine for a scene and not fine for fine print.

Does the transition come with sound?

Yes, and it matters more in this format than in any other. 32 kHz stereo is generated in the same pass as the picture, so the transient lands on the frame the change lands. Name the swell, the hit and the room tone; leave the music unspecified and the model scores it. There is no stock licence to buy and no sync pass.

Does a second aspect ratio cost more?

No. 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16 all bill at the same per-second rate — an extra cut quoted at $500–$1,500 at a studio is just another generation here. Set it on the ratio chip; writing “9:16” into the prompt does nothing, which the car ad on this page demonstrates.

Can I run a generated before and after as an ad?

From our side, yes: commercial rights on every tier and no watermark. From the regulator’s side it is still a performance claim, and a depicted result needs to be one a typical customer actually gets. Health, weight and cosmetic transformations are the sharpest case and should not be generated at all. Platforms additionally require realistic synthetic media to be labelled — Meta, TikTok and YouTube each have the switch in the upload flow.

Why not use a transition pack in my editor?

Because a transition pack joins two assets and the audience can tell. A wipe, a swipe or a split-screen reads as two things filmed separately, which is the one reading a before-and-after cannot survive. What you get here is a single continuous take in which the same frame changes, with the sound of it changing generated on the same pass. Worse if you want a hundred presets; better if you want the join to be invisible.

Show the change, not two photographs of it.

Name what stays still, name the frame where both states look alike, and get one unbroken take back with its own sound. Your first one is free — no card, no watermark, and it is yours to run commercially.

Sign up free · 1 clip on MiniMax H3

Written and maintained by the MiniMax H3 AI Video Generator editorial teamPublished Last updated Tool version 2026.09.4