Aisma vs 7 popular AI video editors:
same clip, same request

We gave Aisma and 7 popular AI video editors the same everyday edits people search for, on real clips: removing people, changing the setting, changing an outfit, swapping a product. Here is each editor’s result, its score and what the edit cost.

Choose your video

Paid editing · MP4 or MOV · 2–60 seconds · Review the price before generating.

See the results
everyday edits
11
editors, Aisma included
8
videos side by side
54
Every result scored blind on four criteria.

AT A GLANCE

Six reasons Aisma
came out ahead.

Close scores cost more.

Wherever another editor came within half a point of Aisma, its edit cost more: on the overshirt, Hailuo AI scored 4.85 to Aisma’s 5.00 at ≈$1.12 against ≈$0.80, and on the dog’s raincoat, Kling AI scored 4.70 to Aisma’s 4.90 at ≈$3.60 against ≈$2.20. Compare the prices side by side.

43 seconds in one go.

On the 43-second winter clip, Aisma turned all 43 seconds to winter in one go. Runway, Kling AI and Higgsfield take up to 30 seconds in one job, Adobe Firefly 20 and Hailuo AI 15, and Google Flow edits up to 15 seconds at a time.

Length, frame and cuts kept.

Every Aisma result came back at its clip’s exact length and in its frame, and on the eight edits with several shots it kept all 23 cuts in place.

No made-up sound.

Every clip was silent. Kling AI, Hailuo AI and Google Flow added generated sound to all their results; Aisma kept every clip silent, as filmed.

One shot changed, one shot priced.

On the dress clip, with the short request and the long one, only one shot of three needed a change, and Aisma priced that shot alone: ≈$0.60, where the next-lowest price for the same edit was ≈$2.10. The others price by length; the price table has every figure.

HOW WE COMPARED

Scored blind,
priced per edit.

Every editor got the same real clip and the same plain request, in its own edit mode.

Every result was scored blind on four criteria: the asked change (30%), the rest kept as filmed (30%), holding steady over time (20%) and looking real (20%).

We also checked each result against its clip: its length, its frame, its sound and, where the clip has several shots, its cuts.

Each price is what that edit cost at 720p, or at the size an editor priced where it has no 720p price, in approximate US dollars.

The seven editors, with the model and mode their apps showed

Runway
Aleph 2.0, Edit Studio
Kling AI
Kling 4.0, Omni reference
Hailuo AI
MiniMax H3, Omni Reference
Google Flow
Omni 1.2 Flash, agent edit
Higgsfield
Genjutsu, Objects swap
Adobe Firefly
Firefly Video 2
Pika
Swap Anything

RESULTS BY EDIT

Side by side,
edit by edit.

The original comes first, then Aisma, then the other editors by score. Tap a clip to play it; one plays at a time, and clips with sound start muted.

The scores, checks and prices in one file (CSV)

EDIT 1

Remove the people,
five shots.

“Remove the people in the background.”

27.5 seconds · 5 shots · no photo · from Pexels: 5375692, 5375693, 5375690, 5375695, 5375696

First frame of the original clip, edit 1
Original27.50 s · silent
First frame of Aisma’s result, edit 1
Aisma27.50 s · silent
First frame of Runway’s result, edit 1
Runway · Aleph 2.0, Edit Studio27.50 s · silent
First frame of Kling AI’s result, edit 1
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI28.04 s · generated sound, starts muted
Blind scores, checks and prices, edit 1
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedCuts keptPrice per edit at 720p
Aisma5.005.05.05.05.0SameYesNo4 of 4≈$2.40
Runway2.901.03.04.04.5SameYesNo4 of 4≈$8.04
Kling AIGenerated by Kling AI2.201.02.02.54.0+0.54 sYesYes2 of 4≈$4.80
Hailuo AINo result: Omni Reference takes clips of 2–15 seconds; this one is 27.5 seconds.
Google FlowNo result: Flow edits up to 15 seconds at a time; this one is 27.5 seconds.
HiggsfieldNo result: Objects swap needs a reference photo, and this request has none.
Adobe FireflyNo result: Firefly Video 2 takes clips of 5–20 seconds; this one is 27.5 seconds.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 1

Aisma

Removed the background people in every shot: the beach pair and a figure on the rocks (8.00, 10.75 s), the walker (16.25–21.75 s) and the seated people (22.25–27.25 s), with clean fills, and left shots 1 and 3 as filmed. Aisma did it in one run plus one extra edit of the last shot (22–27.5 s): “Also remove the shirtless man sitting on the sand on the right, beside the blue and yellow striped beach chair.” The price includes both.

Runway

Removed no one: the beach pair (8.00, 10.75 s), the walker (16.25–21.75 s) and the seated people (22.25–27.25 s) all remain. The sky and light of 11.25–15.75 s were regraded cooler and the subject of shot 5 brightened, which the request did not ask for.

Kling AI

Removed no one: the beach pair (8.00, 10.75 s) and the seated people (24.75 s) remain, and the walker turned into a dark silhouette (19.00 s). Shot 4 was regraded darker, toward night. Its cuts came late: at 5.25, 16.25 and 22.25 s the previous shot still shows.

How to remove someone from a video

EDIT 2

Remove the people,
four shots.

“Remove the people in the background.”

26.5 seconds · 4 shots · no photo · from Pexels: 12185606, 12185612, 12185607, 12185609

First frame of the original clip, edit 2
Original26.50 s · silent
First frame of Aisma’s result, edit 2
Aisma26.50 s · silent
First frame of Runway’s result, edit 2
Runway · Aleph 2.0, Edit Studio26.50 s · silent
First frame of Kling AI’s result, edit 2
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI27.04 s · generated sound, starts muted
Blind scores, checks and prices, edit 2
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedCuts keptPrice per edit at 720p
Aisma4.354.04.54.54.5SameYesNo3 of 3≈$2.10
Runway3.251.54.04.04.0SameYesNo3 of 3≈$7.76
Kling AIGenerated by Kling AI2.351.02.53.03.5+0.54 sYesYes2 of 3≈$4.63
Hailuo AINo result: Omni Reference takes clips of 2–15 seconds; this one is 26.5 seconds.
Google FlowNo result: Flow edits up to 15 seconds at a time; this one is 26.5 seconds.
HiggsfieldNo result: Objects swap needs a reference photo, and this request has none.
Adobe FireflyNo result: Firefly Video 2 takes clips of 5–20 seconds; this one is 26.5 seconds.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 2

Aisma

Cleared the background people in shots 1, 3 and 4 with clean fills. A few tiny distant pedestrians remain (0.25, 6.75 s), with a figure at the far left (16.75, 19.75 s) and a faint one at the right edge (26.25 s).

Runway

Removed at most the foreground cyclist in shot 1, where the cyclists may only have been reshuffled. The pair and café crowd (3.5, 6.75 s), the bench group (13.75–19.75 s) and the walkers and seated people (20.25–26.25 s) remain.

Kling AI

Removed no one in any shot. At 20.25 s it still shows shot 3 where the source has cut to shot 4. At 23.25 and 26.25 s the man in shot 4 has a second backpack on his shoulder beside the one in his hand.

How to remove someone from a video

EDIT 3

Make it night,
four shots.

“Make it night.”

21.75 seconds · 4 shots · no photo · from Pexels: 27902012, 27902034, 27902016, 27902013

First frame of the original clip, edit 3
Original21.75 s · silent
First frame of Aisma’s result, edit 3
Aisma21.75 s · silent
First frame of Runway’s result, edit 3
Runway · Aleph 2.0, Edit Studio21.75 s · silent
First frame of Kling AI’s result, edit 3
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI22.04 s · generated sound, starts muted
Blind scores, checks and prices, edit 3
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedCuts keptPrice per edit at 720p
Aisma4.654.55.04.54.5SameYesNo3 of 3≈$2.20
Runway4.255.03.54.54.0SameYesNo3 of 3≈$6.36
Kling AIGenerated by Kling AI2.352.52.02.03.0+0.29 sYesYes3 of 3≈$3.77
Hailuo AINo result: Omni Reference takes clips of 2–15 seconds; this one is 21.75 seconds.
Google FlowNo result: Flow edits up to 15 seconds at a time; this one is 21.75 seconds.
HiggsfieldNo result: Objects swap needs a reference photo, and this request has none.
Adobe FireflyNo result: Firefly Video 2 takes clips of 5–20 seconds; this one is 21.75 seconds.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 3

Aisma

Turned all four shots into a convincing night with street lights, keeping the people, cars, framing and cuts. The station shot (11.00–16.50 s) is lit a little flatly, closer to dim daylight.

Runway

Turned every shot to night, with lit windows and streetlights. The silver car turned black at 16.50 s and parked cars were removed at 19.25 and 21.50 s. The station shot (11.00–16.50 s) went dark enough to lose background buildings, people and detail.

Kling AI

Shot 2 (6.00–10.50 s) stays in daylight, and from 13.75 to 16.50 s only the sky turned black over a sunlit scene. Shot 2 runs past its cut, still showing at 11.00 s. It added a walker (8.25 s) and a minivan (19.25 s).

How to change the background of a video

EDIT 4

A red raincoat
on a dog, six shots.

“Put a red raincoat on the dog.”

21 seconds · 6 shots · no photo · from Pexels: 8322555, 8322546, 8322559, 8322563, 8322560, 8322550

First frame of the original clip, edit 4
Original21.00 s · silent
First frame of Aisma’s result, edit 4
Aisma21.00 s · silent
First frame of Kling AI’s result, edit 4
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI21.04 s · generated sound, starts muted
First frame of Runway’s result, edit 4
Runway · Aleph 2.0, Edit Studio21.00 s · silent
Blind scores, checks and prices, edit 4
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedCuts keptPrice per edit at 720p
Aisma4.905.05.04.55.0SameYesNo5 of 5≈$2.20
Kling AIGenerated by Kling AI4.705.04.05.05.0+0.04 sYesYes3 of 5≈$3.60
Runway4.204.05.03.54.0SameYesNo5 of 5≈$6.08
Hailuo AINo result: Omni Reference takes clips of 2–15 seconds; this one is 21 seconds.
Google FlowNo result: Flow edits up to 15 seconds at a time; this one is 21 seconds.
HiggsfieldNo result: Objects swap needs a reference photo, and this request has none.
Adobe FireflyNo result: Firefly Video 2 takes clips of 5–20 seconds; this one is 21 seconds.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 4

Aisma

Put a glossy red raincoat on the dog in all six shots, keeping its red collar, the cuts and the framing, with no visible artifacts. The coat’s design may shift a little between shots: hooded at 0.25 s, plainer and without a hood at 10.25 s.

Kling AI

Put the same red hooded raincoat on the dog in every shot. Two cuts come late: at 14.25 s it still shows the walking shot where the source has cut to the man, and at 17.25 s it still shows the bench shot.

Runway

Put a red garment on the dog in every shot, but it reads more as a padded vest than a raincoat. Its design changes between shots: a hooded zip at 6.75 s, a flat band at 12.00 s that looks pasted on and a zip vest at 14.25 s; the hood appears at 6.75 and 17.25 s but not in shot 1.

How to change clothes in a video

EDIT 5

Make it winter,
one 43-second shot.

“Make it winter.”

43 seconds · one shot · no photo · from Pexels: 37909698

First frame of the original clip, edit 5
Original43.00 s · silent
First frame of Aisma’s result, edit 5
Aisma43.00 s · silent
Blind scores, checks and prices, edit 5
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedPrice per edit at 720p
Aisma4.005.03.04.04.0SameYesNo≈$4.90
RunwayNo result: Edit Studio takes clips up to 30 seconds; this one is 43 seconds.
Kling AINo result: Omni reference takes clips of 3–30 seconds; this one is 43 seconds.
Hailuo AINo result: Omni Reference takes clips of 2–15 seconds; this one is 43 seconds.
Google FlowNo result: Flow edits up to 15 seconds at a time; this one is 43 seconds.
HiggsfieldNo result: Objects swap takes clips of 4–30 seconds; this one is 43 seconds.
Adobe FireflyNo result: Firefly Video 2 takes clips of 5–20 seconds; this one is 43 seconds.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 5

Aisma

Turned all 43 seconds to winter in one go, with snow everywhere. It also added scenery the request did not ask for: power lines from 23.86 s, large trees at 33.31 and 42.75 s and a sign at 9.69 s.

How to change the background of a video

EDIT 6

One dress of three,
in a short request.

“Make the white dress red.”

14.79 seconds · 3 shots · no photo · from Pexels: 12723362Edit 7, next, puts a long, detailed request on the same clip.

First frame of the original clip, edit 6
Original14.79 s · silent
First frame of Aisma’s result, edit 6
Aisma14.79 s · silent
First frame of Kling AI’s result, edit 6
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI15.04 s · generated sound, starts muted
First frame of Hailuo AI’s result, edit 6
Hailuo AI · MiniMax H3, Omni Reference15.08 s · generated sound, starts muted
First frame of Runway’s result, edit 6
Runway · Aleph 2.0, Edit Studio14.79 s · silent
First frame of Adobe Firefly’s result, edit 6
Adobe Firefly · Firefly Video 210.08 s · silent
Blind scores, checks and prices, edit 6
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedCuts keptPrice per edit at 720p
Aisma4.605.04.05.04.5SameYesNo2 of 2≈$0.60
Kling AIGenerated by Kling AI4.305.03.05.04.5+0.25 sYesYes2 of 2≈$2.57
Hailuo AI3.204.02.03.04.0+0.29 sYesYes0 of 2≈$2.10
Runway2.151.01.53.53.5SameYesNo2 of 2≈$4.40
Adobe Firefly2.104.01.01.02.0−4.71 sYesNo0 of 2≈$2.63
Google Flow0Failed: Flow removed its own result (“Video removed”).
HiggsfieldNo result: Objects swap needs a reference photo, and this request has none.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 6

Aisma

Turned the white dress red throughout its shot and left the green and the yellow dress shots untouched. In that shot the background turned slightly cooler, and the eyelet lace may read bolder than filmed (3.83, 6.50 s).

Kling AI

Turned the white dress red throughout shot 2. At 3.83 and 6.50 s her tinted glasses became black sunglasses, and the lace dress became plain satin with a flared skirt.

Hailuo AI

Turned shot 2 red but re-staged it: at 3.83 s she faces away. The white close-up at 9.17 s is gone, with the striped-dress shot in its place, and shot 3 (9.67–14.54 s) was regenerated with different framing, poses and cut point.

Runway

Turned the green dress of shot 1 red (0.25–3.33 s) and left the white dress white (3.83, 6.50, 9.17 s). The close-up at 9.17 s is softened or smeared, and the pattern at 9.67 s was recolored.

Adobe Firefly

Turned the dress red in shot 2 (3.83, 6.50 s); at 9.17 s the close-up is smeared and the lace lost. At 9.67 s, where the source shows the blue-and-yellow dress, it still shows a blurry red close-up, and its file ends at 10.08 seconds, so the rest of that shot is missing.

How to change a shirt color in a video

EDIT 7

One dress of three,
in a long request.

“In the shot where she wears the white strapless dress, change it to a solid red dress with the same cut, fit and slits. Keep the shot in the green dress and the shot in the blue-and-yellow striped dress exactly as filmed, and keep her face, hair, sunglasses, sandals, movement, the white wall, the light and the camera.”

14.79 seconds · 3 shots · no photo · from Pexels: 12723362The same clip as edit 6, with a long, detailed request.

First frame of the original clip, edit 7
Original14.79 s · silent
First frame of Aisma’s result, edit 7
Aisma14.79 s · silent
First frame of Kling AI’s result, edit 7
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI15.04 s · generated sound, starts muted
First frame of Hailuo AI’s result, edit 7
Hailuo AI · MiniMax H3, Omni Reference15.08 s · generated sound, starts muted
First frame of Pika’s result, edit 7
Pika · Swap Anything14.71 s · silent
First frame of Adobe Firefly’s result, edit 7
Adobe Firefly · Firefly Video 210.08 s · silent
First frame of Runway’s result, edit 7
Runway · Aleph 2.0, Edit Studio14.79 s · silent
Blind scores, checks and prices, edit 7
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedCuts keptPrice per edit at 720p
Aisma4.705.04.05.05.0SameYesNo2 of 2≈$0.60
Kling AIGenerated by Kling AI2.953.03.52.03.0+0.25 sYesYes1 of 2≈$2.57
Hailuo AI2.853.52.02.53.5+0.29 sYesYes1 of 2≈$2.10
Pika2.803.01.04.04.0−0.08 sYesNo1 of 2≈$4.85, Pika’s estimate
Adobe Firefly2.153.51.02.02.0−4.71 sYesNo1 of 2≈$2.63
Runway2.101.01.03.54.0SameYesNo2 of 2≈$4.40
Google Flow0Failed: Flow showed an error and returned no video.
HiggsfieldNo result: Objects swap needs a reference photo, and this request has none.

What each editor did on edit 7

Aisma

Turned the white dress red throughout its shot, keeping the cut, the slits and the eyelet texture, and kept the green and the striped dress shots as filmed. The wall in that shot turned slightly cooler (3.83–6.50 s).

Kling AI

At 3.83 s the white-dress shot opens in a green dress with a red flap. At 6.50 s the dress is red but flared, with several slits, losing the straight cut. The striped shot may also have been regenerated, with her in another position (12.10, 14.54 s).

Hailuo AI

At 3.83 s she is seen from the back in a green dress, not a red one, and her pose changed at 3.33 s. The striped shot was regenerated with a different pose, position, pattern and motion (12.10–14.54 s), and her face is smeared at 14.54 s.

Pika

Swap Anything needs a reference photo for an outfit, so it also got a photo of a red strapless gown, with the request and a pointer to that photo. It put the same long red satin gown in all three shots, replacing the green dress (0.25–3.33 s) and the striped one (9.67–14.54 s) as well, and changed the cut and length from midi to floor length.

Adobe Firefly

Turned the dress solid red with its slits at 3.83–6.50 s. From 9.17 s the picture goes dark, nearly black at 9.67 s, where the source shows the striped dress, and its file ends at 10.08 seconds, so the rest of that shot is missing. Her pose also appears changed at 3.33 s.

Runway

Turned the green dress red instead (0.25–3.33 s) and left the white dress white (3.83–9.17 s). It put dark sunglasses in every shot, replacing her tinted pair at 6.50 s and adding a new pair at 12.10 s.

How to change clothes in a video

EDIT 8

A can from a photo,
three shots.

“Swap the drink for the can in the photo.”

19.5 seconds · 3 shots · a photo of a white can · from Pexels: 8533907, 8533909, 8533915

First frame of the original clip, edit 8
Original19.50 s · silent
First frame of Aisma’s result, edit 8
Aisma19.50 s · silent
First frame of Higgsfield’s result, edit 8
Higgsfield · Genjutsu, Objects swap19.38 s · silent
First frame of Kling AI’s result, edit 8
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI20.04 s · generated sound, starts muted
First frame of Runway’s result, edit 8
Runway · Aleph 2.0, Edit Studio19.50 s · silent
Blind scores, checks and prices, edit 8
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedCuts keptPrice per edit at 720p
Aisma4.555.04.54.04.5SameYesNo2 of 2≈$2.10
Higgsfield3.953.55.03.53.5−0.12 sYesNo2 of 2≈$6.77
Kling AIGenerated by Kling AI3.703.05.03.53.0+0.54 sYesYes2 of 2≈$3.43
Runway2.401.52.52.04.0SameYesNo2 of 2≈$5.80
Hailuo AINo result: Omni Reference takes clips of 2–15 seconds; this one is 19.5 seconds.
Google FlowNo result: Flow edits up to 15 seconds at a time; this one is 19.5 seconds.
Adobe FireflyNo result: This request comes with a photo, and Firefly Video 2 takes no reference photo.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 8

Aisma

Replaced the bottle with the white can in all three shots and kept the scene. The grip on it changes at 3.25 s, the framing may shift a little at 6.25 s, and in shot 3 the can turns light gray (13.25, 16.25 s).

Higgsfield

Put a white can in shots 1 and 2, though the object at 6.25 s may read as an unclear tall cylinder. In shot 3 (13.25–16.25 s) it is an oversized white cylinder with the label removed and no can rim, closer to the bottle or a tube than a can.

Kling AI

Swapped in a can in every shot, but a chrome blue or silver one, not the white can from the photo. At 6.25 and 12.75 s the can keeps the sport bottle’s cap, and its color shifts between shots.

Runway

Shows the can only at 0.25 s; from 3.25 s on the bottle stays. It added text to the bottle at 6.25 s and replaced the stadium seats and trees of shot 3 with track (13.25, 19.25 s).

How to swap a product in a video

EDIT 9

A jacket from a photo,
three shots.

“Put him in the jacket from the photo.”

14 seconds · 3 shots · a photo of a black leather jacket · from Pexels: 8151834, 8151840

First frame of the original clip, edit 9
Original14.00 s · silent
First frame of Aisma’s result, edit 9
Aisma14.00 s · silent
First frame of Kling AI’s result, edit 9
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI14.04 s · generated sound, starts muted
First frame of Higgsfield’s result, edit 9
Higgsfield · Genjutsu, Objects swap13.71 s · silent
First frame of Runway’s result, edit 9
Runway · Aleph 2.0, Edit Studio14.00 s · silent
First frame of Hailuo AI’s result, edit 9
Hailuo AI · MiniMax H3, Omni Reference14.38 s · generated sound, starts muted
Blind scores, checks and prices, edit 9
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedCuts keptPrice per edit at 720p
Aisma5.005.05.05.05.0SameYesNo2 of 2≈$1.40
Kling AIGenerated by Kling AI4.754.55.05.04.5+0.04 sYesYes2 of 2≈$2.40
Higgsfield4.604.05.05.04.5−0.29 sYesNo2 of 2≈$4.74
Runway4.404.05.04.04.5SameYesNo2 of 2≈$4.12
Hailuo AI3.204.02.03.04.0+0.38 sYesYes1 of 2≈$1.96
Google Flow0Failed: Flow refused the clip upload, so nothing was generated.
Adobe FireflyNo result: This request comes with a photo, and Firefly Video 2 takes no reference photo.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 9

Aisma

Put him in a black, distressed leather jacket with a stand collar and zip pockets, a close match for the photo, in all three shots. His face, pose and red trousers, the background and the cuts were kept, with no visible problems.

Kling AI

Put him in a black leather jacket with zip pockets in all shots and kept the scene, his face and the cuts. The jacket is close to the photo, though it may be smoother, with a more generic collar than the photo’s distressed stand collar. His right hand on the railing is smeared at 4.75 s.

Higgsfield

Put him in a distressed stand-collar leather jacket with zip pockets in every shot, but faded charcoal-gray rather than the photo’s black. At 7.50 s his raised hand is blurry and its pose differs slightly; the scene was kept.

Runway

Put him in a glossy leather jacket, not the photo’s distressed one, with few of its zip details. The design changes between shots: a shirt collar at 0.25 and 12.00 s, a stand collar and flap pockets at 7.50 s. His right hand is slightly smeared at 4.75 s.

Hailuo AI

Put him in a jacket close to the photo in every shot, but changed the shots: shot 2 already shows at 4.75 s, his pose changed at 7.50 s, and shot 3 (9.75–13.75 s) was replaced by a different canopy-pier scene and framing.

How to change clothes in a video

EDIT 10

A new overshirt,
one 8-second shot.

“Change his plaid overshirt to a plain camel corduroy overshirt with the same fit. Keep his brown beanie, beige hoodie, face, movement, the boardwalk, the water, the light and the camera.”

7.96 seconds · one shot · no photo · from Pexels: 7496173

First frame of the original clip, edit 10
Original7.96 s · silent
First frame of Aisma’s result, edit 10
Aisma7.96 s · silent
First frame of Hailuo AI’s result, edit 10
Hailuo AI · MiniMax H3, Omni Reference8.00 s · generated sound, starts muted
First frame of Runway’s result, edit 10
Runway · Aleph 2.0, Edit Studio7.96 s · silent
First frame of Kling AI’s result, edit 10
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI8.04 s · generated sound, starts muted
First frame of Google Flow’s result, edit 10
Google Flow · Omni 1.2 Flash, agent edit8.04 s · generated sound, starts muted
Blind scores, checks and prices, edit 10
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedPrice per edit at 720p
Aisma5.005.05.05.05.0SameYesNo≈$0.80
Hailuo AI4.855.04.55.05.0+0.04 sNoYes≈$1.12
Runway4.405.03.05.05.0SameYesNo≈$2.44
Kling AIGenerated by Kling AI4.404.04.05.05.0+0.08 sYesYes≈$1.37
Google Flow4.405.03.05.05.0+0.08 sNoYes≈$0.48
HiggsfieldNo result: Objects swap needs a reference photo, and this request has none.
Adobe FireflyNo result: Firefly Video 2 grayed itself out for this clip: “Not compatible with your current video”.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 10

Aisma

Changed the overshirt to camel corduroy at every moment and kept the red inner layer, the bag strap, the hoodie, the beanie, his face and the boardwalk. Clean and consistent throughout, with no visible problems.

Hailuo AI

Changed the overshirt to camel corduroy with a similar fit at every moment and kept the strap, the red inner layer, the hoodie, the beanie, his face and the boardwalk. At 6.22 and 7.71 s his head pose and eyes may drift slightly, with a small shift in framing.

Runway

Changed the overshirt to camel corduroy throughout, but removed the red inner layer and the crossbody strap at every moment (0.25, 6.22 s); the hoodie fills the gap.

Kling AI

Changed the overshirt to corduroy throughout, but in a darker rust or cognac rather than camel, and removed the bag strap at every moment (0.25–7.71 s). It kept the red inner layer, which may look repatterned at 6.22 s, the hoodie, the beanie and the scene.

Google Flow

Changed the overshirt to camel corduroy everywhere, with a slightly boxier fit, but replaced the red inner layer and removed the crossbody strap throughout (0.25–7.71 s).

How to change clothes in a video

EDIT 11

Remove a car,
one 8-second shot.

“Remove the white SUV from the driveway for the whole clip and rebuild the paving and the garage behind it. Keep the houses, the tree, the planter, the gate, the light and the camera.”

7.96 seconds · one shot · no photo · from Pexels: 29635453

First frame of the original clip, edit 11
Original7.96 s · silent
First frame of Aisma’s result, edit 11
Aisma7.96 s · silent
First frame of Kling AI’s result, edit 11
Kling AI · Kling 4.0, Omni referenceGenerated by Kling AI8.04 s · generated sound, starts muted
First frame of Hailuo AI’s result, edit 11
Hailuo AI · MiniMax H3, Omni Reference8.00 s · generated sound, starts muted
First frame of Adobe Firefly’s result, edit 11
Adobe Firefly · Firefly Video 27.96 s · silent
First frame of Runway’s result, edit 11
Runway · Aleph 2.0, Edit Studio7.96 s · silent
First frame of Google Flow’s result, edit 11
Google Flow · Omni 1.2 Flash, agent edit8.04 s · generated sound, starts muted
Blind scores, checks and prices, edit 11
EditorScoreChange (30%)Kept (30%)Time (20%)Real (20%)LengthFrame keptSound addedPrice per edit at 720p
Aisma5.005.05.05.05.0SameYesNo≈$0.80
Kling AIGenerated by Kling AI4.705.04.05.05.0+0.08 sYesYes≈$1.37
Hailuo AI4.705.04.05.05.0+0.04 sYesYes≈$1.12
Adobe Firefly4.705.04.05.05.0SameYesNo≈$2.11
Runway4.505.04.04.54.5SameYesNo≈$2.44
Google Flow4.305.03.05.04.5+0.08 sYesYes≈$0.48
HiggsfieldNo result: Objects swap needs a reference photo, and this request has none.
PikaNo result: Swap Anything is built around reference photos; Pika’s result is on edit 7.

What each editor did on edit 11

Aisma

Removed the SUV at every moment and rebuilt the paving, joints and all, and the open garage with the pickup inside it, as filmed. The houses, tree, planter, gate and camera were kept, with a slight overall shift in brightness.

Kling AI

Removed the SUV throughout and kept the rest of the scene. It rebuilt the garage dark and empty, so the pickup inside it is gone too (6.22–7.71 s).

Hailuo AI

Removed the SUV throughout and kept everything else, but rebuilt the open garage as a closed garage door, which hides the pickup (0.25–7.71 s).

Adobe Firefly

Removed the SUV throughout and kept the rest, but rebuilt the garage interior lighter and empty, removing the pickup inside it (6.22, 7.71 s).

Runway

Removed the SUV throughout and kept everything else, but rebuilt the garage as an empty, lit carport, so the pickup inside is gone (6.22, 7.71 s). Shelves inside the garage appear and shift between 0.25 s and 3.23–6.22 s.

Google Flow

Removed the SUV at every moment, but lit the house windows and doorway with warm lights the request did not ask for (0.25–7.71 s). It repaved the driveway smoother and lighter than the source and removed the pickup in the garage.

How to remove objects from a video

PRICES

What each edit
cost.

Close scores, higher prices: ten results from the other editors came within half a point of Aisma’s score, and every one of those edits cost more than Aisma’s.

What each result’s edit cost at 720p, in approximate US dollars, beside its score
EditEditorScorePrice per edit at 720p
Edit 1Aisma5.00≈$2.40
Edit 1Runway2.90≈$8.04
Edit 1Kling AIGenerated by Kling AI2.20≈$4.80
Edit 2Aisma4.35≈$2.10
Edit 2Runway3.25≈$7.76
Edit 2Kling AIGenerated by Kling AI2.35≈$4.63
Edit 3Aisma4.65≈$2.20
Edit 3Runway4.25≈$6.36
Edit 3Kling AIGenerated by Kling AI2.35≈$3.77
Edit 4Aisma4.90≈$2.20
Edit 4Kling AIGenerated by Kling AI4.70≈$3.60
Edit 4Runway4.20≈$6.08
Edit 5Aisma4.00≈$4.90
Edit 6Aisma4.60≈$0.60
Edit 6Kling AIGenerated by Kling AI4.30≈$2.57
Edit 6Hailuo AI3.20≈$2.10
Edit 6Runway2.15≈$4.40
Edit 6Adobe Firefly2.10≈$2.63
Edit 7Aisma4.70≈$0.60
Edit 7Kling AIGenerated by Kling AI2.95≈$2.57
Edit 7Hailuo AI2.85≈$2.10
Edit 7Pika2.80≈$4.85, Pika’s estimate
Edit 7Adobe Firefly2.15≈$2.63
Edit 7Runway2.10≈$4.40
Edit 8Aisma4.55≈$2.10
Edit 8Higgsfield3.95≈$6.77
Edit 8Kling AIGenerated by Kling AI3.70≈$3.43
Edit 8Runway2.40≈$5.80
Edit 9Aisma5.00≈$1.40
Edit 9Kling AIGenerated by Kling AI4.75≈$2.40
Edit 9Higgsfield4.60≈$4.74
Edit 9Runway4.40≈$4.12
Edit 9Hailuo AI3.20≈$1.96
Edit 10Aisma5.00≈$0.80
Edit 10Hailuo AI4.85≈$1.12
Edit 10Runway4.40≈$2.44
Edit 10Kling AIGenerated by Kling AI4.40≈$1.37
Edit 10Google Flow4.40≈$0.48
Edit 11Aisma5.00≈$0.80
Edit 11Kling AIGenerated by Kling AI4.70≈$1.37
Edit 11Hailuo AI4.70≈$1.12
Edit 11Adobe Firefly4.70≈$2.11
Edit 11Runway4.50≈$2.44
Edit 11Google Flow4.30≈$0.48

How we priced each edit

Each price is that edit’s cost at 720p, or at the size an editor priced where it has no 720p price. Each editor sells credits at its own price; we converted the credits each edit used into dollars at the price we paid for them, Google Flow’s at its US list price and Aisma’s at its Starter plan’s. Prices change often.

Aisma is priced by the second: about $0.10 per counted second at 720p, shot by shot, and parts with nothing to change cost nothing. Pricing shows how the seconds are counted.

WHAT EACH EDITOR TAKES

Aisma takes clips
up to 60 seconds long.

Limits as each editor’s app and help pages stated them. The last column is what came back on the 11 edits, whose clips are all silent; Aisma keeps a clip’s own sound.

What each editor takes, as its app and help pages stated it, and what came back on the 11 edits
EditorLongest clip in one jobReference photosIn our results
Aisma2–60 seconds, up to 500 MBUp to 10Same length and frame; silent, as filmed
Runway · Aleph 2.0, Edit StudioUp to 30 secondsYes, as a reference imageSame length and frame; silent, as filmed
Kling AI · Kling 4.0, Omni reference3–30 seconds a clipYesUp to 0.54 s longer; same frame; generated sound added
Hailuo AI · MiniMax H3, Omni Reference2–15 seconds a clip, 15 seconds in allYesUp to 0.38 s longer; a changed frame on edit 10; generated sound added
Google Flow · Omni 1.2 Flash, agent editUploads up to 60 seconds, trimmed to 30; one edit covers up to 10 seconds, 15 on paid plans—0.08 s longer; a changed frame on edit 10; generated sound added
Higgsfield · Genjutsu, Objects swap4–30 secondsRequired: it sends no job without oneUp to 0.29 s shorter; same frame; silent, as filmed
Adobe Firefly · Firefly Video 25–20 seconds, under 200 MBNoThe first 10.08 seconds of the 14.79-second clip on edits 6 and 7, same length on edit 11; same frame; silent, as filmed
Pika · Swap AnythingShown in the appRequired for an outfit0.08 s shorter; same frame; silent, as filmed

FROM THE FOUNDER

Why we put Aisma
side by side.

Aleksei Babkin, founder of Aisma

Why we ran it

“I want to show how much better our AI understands simple, human editing requests, and how well it handles both long scenes and even a video made up of lots of short scenes.”

Aleksei Babkin, founder of Aisma

What a good edit means

“If an edit worked, the person will see that everything in the video changed the way they wanted and exactly how they wanted it, and everything they wanted to keep stayed the same, with no changes.”

Aleksei Babkin, founder of Aisma

Where AI video editors break on real footage

“Most often, AI can’t properly handle a video where the scenes change a lot. It gets lost about what needs to change in which scene and what should stay. It might put a hat on a person in one scene, then remove an object in another scene and forget about the hat. Also, regular editors often charge for the whole video, even if only a few seconds in one of its shots need to change.”

Aleksei Babkin, founder of Aisma

YOUR NEXT EDIT STARTS HERE

Try the same edits
on your own video.

Choose your video

Paid editing · MP4 or MOV · 2–60 seconds · Review the price before generating.

Tests and scores by the Aisma team. Prices as listed on November 11, 2026, in US dollars at the credit price we paid, Google Flow’s at its US list price and Aisma’s at its Starter plan’s; they change often. Questions: support@aisma.ai.