Home/Benchmark Lab

Benchmark Lab

Seedance 2 Fast Compare Pack

Seven paired benchmark cases built from the compare benchmark pack, each with the original prompt, native clip metadata, structured verdicts, and side-by-side playback.

This is a third-party reference page built from repo assets and Gemini 3 Pro review notes. It is not official benchmark documentation from any vendor.

Cases

7

Videos

14

Written reviews

7

Opponent families

6

April 2026 context

Where Kling 3, Veo 3.1, and Sora fit for live searches

Benchmark cases above are frozen repo pairings. Separately, news-driven SEO clusters in 2026 highlight three poles: Kling 3.0 for high headline specs in third-party tables, Google Veo 3.1 for API + cloud procurement stories, and OpenAI Sora winding down its standalone app/API (verify on OpenAI). Numbers below are orientation only — confirm on each vendor before you buy.

Sora 2 appears in historical benchmark files on this page; major outlets reported OpenAI discontinued the Sora consumer product/API direction in March 2026. Treat Sora rows in old cases as archival pairings, not a live shopping path.

Model / routeNative A/V (public claims)Headline specs in blogsTypical API storySEO / market notes
Seedance 2.0Often described as joint audio+video generation in product copy.Up to ~15s, 2K-class positioning in official materials.BytePlus / ModelArk-style paths in enterprise posts; consumer via Dreamina/CapCut in select regions.Strong multimodal @reference story; phased regional rollout.
Kling 3.0Frequently listed alongside other flagship models with synchronized audio in roundup posts.Fourth-gen blog tables often emphasize resolution/FPS leadership claims.Third-party API hosts quote per-second pricing; verify with Kuaishou or authorized partners.Common comparison target vs Seedance in 2026 leaderboards.
Google Veo 3.1Positioned as audio-native cinematic generation in Google ecosystem posts.Benchmark posts vary by tier; prioritize Google Cloud docs.Strong when Gemini / Vertex distribution already approved internally.Frequent “official API” framing vs packaged creator apps.
OpenAI Sora 2 — discontinued path (press)Historically marketed for cinematic clips when the program was active.Archival benchmark pairings only on this page.Reported wind-down of standalone Sora app/API in March 2026 — confirm on OpenAI.High-volume “Sora alternative” searches after announcement.
Pika 2.x-class routesVaries by product generation; many shorts-first tools add audio in post.Community lore emphasizes stylized motion + fast turnaround in blog roundups.Often creator-app first; verify whether batch/API meets your SLA.Frequent “fast meme clip” comparisons beside Seedance/Kling in trend-chasing SEO.

This grid is editorial segmentation for researchers, not a scorecard. Compare Benchmark cases remain the authoritative on-page experiment; the grid tracks why people search in April 2026.

How This Page Reads

Each case keeps the same index, reference frame, and benchmark prompt. The page surfaces the recorded pairing, native playback metadata, and the stored review verdict from the repo.

Repo-backed source

Case pairings, scores, and verdicts are pulled from compare/case-manifest.json.

Native playback

The page preserves original duration, fps, and resolution instead of normalizing clips.

Single review basis

Current verdicts come from Gemini 3 Pro review passes and are stored case by case.

Current Coverage

7 / 7

The current pack covers character motion, cloud control, fluid physics, aerial drift, off-road action, fashion walking, and dense city traffic.

Videos and case frames are streamed from the compare benchmark asset path configured for this site.

Model Navigation

Jump by compared model

Use these model cards to open the matching benchmark cases first, then branch into the linked image and video generators.

SR

Cases

Sora 2 (discontinued Mar 2026, press)

1

Seedance 2 Fast vs Sora 2 (discontinued Mar 2026, press)

Case 1

Generator buttons open external tools on dev.elser.ai.

Generate a similar video on Elser

Generate a similar AI video or image on Elser, then compare your output against this benchmark pairing.

Case Filter

View only one opponent model at a time

Use the filter to reduce the page to one compared model, then inspect only the matched cases and verdict cards below.

Active filter: All compared models

Benchmark Summary

Seven case verdicts, one quick navigation layer

Use this summary rail to jump straight into each paired case, or reuse the same frame and prompt in the linked image and video generators.

Generator buttons open external tools on dev.elser.ai in a new tab. The case card keeps the compared model, winning verdict, and repo-backed conclusion together.

Cliff runner vs Sora 2 summary frame

Case 1

Winner: Sora 2 (discontinued Mar 2026, press)

Seedance 2 Fast vs Sora 2 (discontinued Mar 2026, press)

Sora 2 wins because it keeps the runner, fabric, and environment more stable while still holding the shot language over a longer clip.

External generator tools open on dev.elser.ai.

Generate a similar video on Elser

Generate a similar AI video or image on Elser, then compare your output against this benchmark pairing.

Open case
Glass bridge vs Wan 2.6 Flash summary frame

Case 2

Winner: Wan 2.6 Flash

Seedance 2 Fast vs Wan 2.6 Flash

Wan 2.6 Flash wins because it holds the cloud mass around the bridge and treats the cloud motion more like breathing than directional travel.

External generator tools open on dev.elser.ai.

Generate a similar video on Elser

Generate a similar AI video or image on Elser, then compare your output against this benchmark pairing.

Open case
Balloon drift vs Kling 3 summary frame

Case 4

Winner: Kling 3

Seedance 2 Fast vs Kling 3

Kling 3 wins because it actually executes the split trajectories across depth layers instead of relying on a mostly static scenic plate.

External generator tools open on dev.elser.ai.

Generate a similar video on Elser

Generate a similar AI video or image on Elser, then compare your output against this benchmark pairing.

Open case
London fashion walk vs Seedance Pro Fast summary frame

Case 6

Winner: Seedance Pro Fast

Seedance 2 Fast vs Seedance Pro Fast

Seedance Pro Fast wins as a clear version upgrade because the coat, boots, bag, and walking rhythm remain far more stable over time.

External generator tools open on dev.elser.ai.

Generate a similar video on Elser

Generate a similar AI video or image on Elser, then compare your output against this benchmark pairing.

Open case
Downtown traffic vs Veo 3.1 Fast summary frame

Case 7

Winner: Veo 3.1 Fast

Seedance 2 Fast vs Veo 3.1 Fast

Veo 3.1 Fast wins because it treats the city as a live simulation with vehicle-level decisions instead of a mostly animated plate.

External generator tools open on dev.elser.ai.

Generate a similar video on Elser

Generate a similar AI video or image on Elser, then compare your output against this benchmark pairing.

Open case

Case 1

Winner: Sora 2 (discontinued Mar 2026, press)

Cliff runner vs Sora 2

Checks lateral tracking, running rhythm, cloth motion, and background stability on a sunset plateau.

Sora 2 wins because it keeps the runner, fabric, and environment more stable while still holding the shot language over a longer clip.

Reference frame

Cliff runner vs Sora 2 frame

Original prompt

Prompt text is kept as the original benchmark source text.

A young woman in a flowing white dress runs gracefully across a long, wide cliff plateau at sunset.

The cliff surface is flat and extended - a safe, continuous rocky plain designed for running.

She runs smoothly forward along the cliff edge, never approaching the drop.

Camera tracks her from the side at medium distance, keeping her side profile clearly visible at all times.

Her hair streams backward naturally in the wind.
Her dress flows softly with each step.

Natural running rhythm - light barefoot steps, steady breathing.

She maintains forward momentum in one direction only.

No stopping.
No turning back.
No approaching the cliff edge.

The background reveals vast mountain valleys under warm golden-hour light.

Soft wind moves grass subtly.

Subtle dust lifts from her footsteps.

Camera movement:
- smooth lateral tracking
- slight forward drift
- gentle handheld cinematic motion

Lighting:
golden sunset rim light on her face and body,
soft shadows across the plateau.

Motion style:
realistic running speed,
elegant posture,
relaxed arms,
natural stride.

Atmosphere:
open, cinematic, emotional, freedom-inspired.

No teleporting.
No cuts.
No scene changes.

End with her still running forward, centered in frame, dress flowing behind her.

Ultra-realistic cinematic outdoor fashion style.
Natural physics.
High-end travel film mood.

Side A

Seedance 2 Fast

Total score

65 / 100

39 / 60

Duration

5.07s

FPS

30

Resolution

1112x836

Side B

Sora 2 (discontinued Mar 2026, press)

Total score

83 / 100

50 / 60

Duration

4.04s

FPS

24

Resolution

1176x784

Normalized score table

MetricSeedance 2 FastSora 2 (discontinued Mar 2026, press)
Prompt fit79
Motion / action68
Physics57
Camera89
Stability68
Cinematic finish79

Seedance 2 Fast readout

Seedance 2 Fast matches the visual mood well, but leg deformation and mountain flicker reduce reliability.

Winning readout

Sora 2 keeps stronger detail, light handling, and temporal stability, even though some foot sliding still appears.

Fairness note

Clip lengths differ, but the longer Sora sample remains more stable, so the verdict still leans clearly toward B.

Case 2

Winner: Wan 2.6 Flash

Glass bridge vs Wan 2.6 Flash

Checks whether clouds stay around the bridge instead of drifting away, while reflections and slow walking remain natural.

Wan 2.6 Flash wins because it holds the cloud mass around the bridge and treats the cloud motion more like breathing than directional travel.

Reference frame

Glass bridge vs Wan 2.6 Flash frame

Original prompt

Prompt text is kept as the original benchmark source text.

Single continuous shot.

A person walks slowly forward across a high-altitude glass suspension bridge.

Large cloud masses remain surrounding the bridge and valley.

Clouds do NOT travel away.

They remain present in frame at all times.

Cloud motion is minimal and localized:

The entire cloud body shifts slightly sideways by a small distance,
then stabilizes again.

Cloud edges softly expand and contract, like slow breathing.

No sweeping drift.
No rolling layers.
No disappearing clouds.

Mist curls gently along cliff faces, then settles.

Cloud volume stays consistent.

Only subtle positional change and soft shape morphing.

Sunlight diffuses through the clouds, creating slow moving highlights on the glass panels.

Reflections change gently.

Bridge cables respond slightly to calm wind.

The person continues walking naturally forward.

Camera follows smoothly from behind.

Atmosphere feels dreamy, romantic, peaceful.

Natural physics only.

End with clouds still surrounding the bridge, softly hovering.

Side A

Seedance 2 Fast

Total score

72 / 100

43 / 60

Duration

5.06s

FPS

60

Resolution

1112x834

Side B

Wan 2.6 Flash

Total score

87 / 100

52 / 60

Duration

5.01s

FPS

30

Resolution

1764x1176

Normalized score table

MetricSeedance 2 FastWan 2.6 Flash
Prompt fit69
Motion / action89
Physics78
Camera89
Stability78
Cinematic finish79

Seedance 2 Fast readout

Seedance 2 Fast keeps the warm tone and centered follow shot, but the cloud bank drifts too far and weakens prompt compliance.

Winning readout

Wan keeps the bridge, reflections, and cloud volume more coherent, which better matches the peaceful mood of the brief.

Fairness note

Durations are close here, so the decision mainly comes from cloud control and bridge optics rather than runtime differences.

Case 3

Winner: Wan 2.6

Impact wave vs Wan 2.6

Checks a locked camera, rock stability, splash weight, and believable runoff after impact.

Wan 2.6 wins because it separates moving water from static rocks much more cleanly and produces heavier, more believable runoff.

Reference frame

Impact wave vs Wan 2.6 frame

Original prompt

Prompt text is kept as the original benchmark source text.

Single continuous cinematic shot.

A massive ocean wave surges forward and violently crashes into rugged coastal rocks.

The wave curls, then explodes on impact - white water and mist burst upward in slow chaotic sheets.

Heavy seawater flows back over the dark rocks in thick streams.

Smaller secondary waves follow immediately after, creating layered motion.

Camera is locked in place, low angle facing the rocks.

No camera push-in.
No zoom.

Only the ocean moves.

Water physics are fully realistic:

- wave crest collapses naturally
- foam spreads outward on impact
- spray particles drift in wind
- gravity pulls water back into the sea
- wet rocks glisten under sunset light

Fine mist floats through the air.

Wind subtly bends the spray sideways.

Golden-hour sunlight passes through airborne droplets, creating glowing highlights.

Background waves continue rolling slowly.

Foreground wave hits hard.

High contrast between dark rocks and bright foam.

Ultra-realistic ocean simulation.
Natural turbulence.
No looping.

Energy builds -> impact -> cascading runoff.

Mood: powerful, cinematic, raw nature.

Strong sense of scale.

5 seconds.

End with residual water flowing down rocks while distant waves keep moving.

Side A

Seedance 2 Fast

Total score

63 / 100

38 / 60

Duration

5.06s

FPS

60

Resolution

1112x834

Side B

Wan 2.6

Total score

90 / 100

54 / 60

Duration

5.01s

FPS

30

Resolution

1764x1176

Normalized score table

MetricSeedance 2 FastWan 2.6
Prompt fit79
Motion / action78
Physics69
Camera610
Stability59
Cinematic finish79

Seedance 2 Fast readout

Seedance 2 Fast gives a dramatic splash and strong sunlit spray, but rock deformation and water flicker break the realism.

Winning readout

Wan keeps the background anchored, the impact layered, and the return flow heavy enough to feel physically grounded.

Fairness note

The two clips are effectively matched in length, so the deciding factor is scene stability and fluid behavior, not runtime.

Case 4

Winner: Kling 3

Balloon drift vs Kling 3

Checks layered balloon motion: foreground rise, mid-distance left drift, and far-right background drift under a moving aerial camera.

Kling 3 wins because it actually executes the split trajectories across depth layers instead of relying on a mostly static scenic plate.

Reference frame

Balloon drift vs Kling 3 frame

Original prompt

Prompt text is kept as the original benchmark source text.

Single continuous cinematic aerial shot.

Early morning sunrise over river canyon.

Multiple hot air balloons already airborne.

IMPORTANT: balloons are NOT static.

Each balloon follows a visible trajectory:

Foreground balloons rise upward continuously (about one basket height during the shot).

Mid-distance balloons drift slowly left.

Far balloons drift slowly right.

All balloons move at different speeds.

No balloon stays fixed relative to landscape.

Camera glides forward gently.

Foreground balloons visibly change position against canyon cliffs.

River reflections slide slowly.

Light valley mist flows horizontally.

Baskets sway slightly under envelopes.

Balloon envelopes breathe subtly with warm air.

Natural wind motion.

No frozen objects.

No static composition.

Real physical drift.

Romantic cinematic travel atmosphere.

Ultra realistic.

End with balloons still drifting.

Side A

Seedance 2 Fast

Total score

62 / 100

37 / 60

Duration

5.07s

FPS

30

Resolution

1112x836

Side B

Kling 3

Total score

87 / 100

52 / 60

Duration

4.04s

FPS

24

Resolution

1280x720

Normalized score table

MetricSeedance 2 FastKling 3
Prompt fit69
Motion / action59
Physics68
Camera79
Stability68
Cinematic finish79

Seedance 2 Fast readout

Seedance 2 Fast looks pretty, but the far balloons barely detach from the background and the motion reads more like pseudo-animation.

Winning readout

Kling gives the balloons buoyancy, basket sway, and distinct drift paths, which makes the canyon feel like a live aerial scene.

Fairness note

Kling runs longer, but that longer clip also sustains many more independent trajectories without falling apart, so the gap remains meaningful.

Case 5

Winner: Kling 3

Motocross chase vs Kling 3

Checks low-angle side tracking, rider posture, sand deformation, suspension behavior, and the SUV follow relationship.

Kling 3 wins because it feels like real off-road cinematography rather than a rider plate sliding across a desert background.

Reference frame

Motocross chase vs Kling 3 frame

Original prompt

Prompt text is kept as the original benchmark source text.

Single continuous cinematic shot.

Ultra-realistic action cinematography.

A motocross rider charges across massive golden desert dunes at real off-road racing speed.

The rider leans naturally into sweeping turns - not over-forced - allowing the bike to glide smoothly over soft sand.

Rear tire throws sand outward in wide arcs, particles flying with gravity-driven falloff.

Behind him, a rugged off-road SUV follows at distance, bouncing rhythmically over dunes, generating rolling dust clouds.

Camera tracks beside and slightly ahead of the rider at low angle, close to ground level.

Forward momentum feels effortless and powerful.

Suspension compresses and rebounds naturally over dune crests.

No struggling motion.
No stiff steering.
The bike flows with terrain.

Sand sprays dynamically but never blocks the lens.

Wind pushes dust trails backward.

Sunlight catches airborne particles, producing glowing highlights.

The rider maintains confident athletic posture:

- relaxed forward lean
- bent elbows
- steady throttle control
- focused gaze

The SUV remains visible through dust in background, headlights shimmering softly.

Camera exhibits subtle terrain vibration.

No slow motion.

Real-world racing speed.

Natural physics:

- realistic wheel spin
- soft dune deformation
- gravity-driven sand falloff
- turbulent dust behavior
- aerodynamic debris trails

Golden hour desert lighting.

Foreground ultra sharp.
Background slight motion blur.

Strong depth of field.

Camera smoothly matches rider speed, holding him centered in frame.

The rider continues accelerating forward while the camera maintains parallel tracking.

Sand washes past the lower frame edge only.

SUV stays chasing in background.

Ending shot:

Rider still charging forward, perfectly framed mid-action.

No subject exits frame.

No impossible motion.

No teleporting.

Camera holds composition until cut.

High kinetic energy.
Authentic off-road riding feel.
Pure cinematic realism.

Side A

Seedance 2 Fast

Total score

53 / 100

32 / 60

Duration

5.07s

FPS

30

Resolution

1112x836

Side B

Kling 3

Total score

90 / 100

54 / 60

Duration

4.04s

FPS

24

Resolution

1280x720

Normalized score table

MetricSeedance 2 FastKling 3
Prompt fit69
Motion / action59
Physics59
Camera710
Stability48
Cinematic finish59

Seedance 2 Fast readout

Seedance 2 Fast nails the golden look and framing, but the rider-sand interaction, turning posture, and temporal stability fall short.

Winning readout

Kling shows convincing wheel bite, sand arcs, suspension compression, and background SUV motion without common structural collapse.

Fairness note

The longer Kling clip carries more risk in a high-speed follow shot, yet it stays cleaner than the shorter Seedance sample, so B still wins clearly.

Case 6

Winner: Seedance Pro Fast

London fashion walk vs Seedance Pro Fast

Checks walking stability, clothing texture, accessory consistency, shallow depth of field, and fashion-ad finish.

Seedance Pro Fast wins as a clear version upgrade because the coat, boots, bag, and walking rhythm remain far more stable over time.

Reference frame

London fashion walk vs Seedance Pro Fast frame

Original prompt

Prompt text is kept as the original benchmark source text.

Stylish British autumn street scene.

A fashionable young woman in full-body view walks confidently forward along a London street covered with fallen leaves.
She wears a classic houndstooth coat, knee-high leather boots, beret, sunglasses, and carries a handbag.

Camera performs a smooth forward tracking shot, slightly low angle, following her stride.

Her hair and coat move naturally in the breeze.

Background shows iconic London architecture, red phone booth, passing cars and pedestrians, creating lively urban motion.

Natural daylight with soft golden autumn tones.

Strong fashion-commercial energy, European luxury street style.

Subtle motion blur on passing traffic.

Realistic walking speed.

High-end cinematic look, shallow depth of field, crisp subject focus.

No pose freezing.
No slow motion.
Continuous movement.

Ultra realistic.

5 seconds.

Side A

Seedance 2 Fast

Total score

60 / 100

36 / 60

Duration

5.07s

FPS

30

Resolution

1112x836

Side B

Seedance Pro Fast

Total score

87 / 100

52 / 60

Duration

5.04s

FPS

24

Resolution

1248x704

Normalized score table

MetricSeedance 2 FastSeedance Pro Fast
Prompt fit79
Motion / action69
Physics68
Camera79
Stability59
Cinematic finish58

Seedance 2 Fast readout

Seedance 2 Fast gets the autumn palette and camera angle right, but leg shape, boot edges, and background structure remain fragile.

Winning readout

Seedance Pro Fast keeps texture, gait, and subject isolation at a level that feels directly reusable in fashion or creator content.

Fairness note

The Pro clip runs longer but still holds more detail and stability, so the extra runtime strengthens the verdict rather than weakening it.

Case 7

Winner: Veo 3.1 Fast

Downtown traffic vs Veo 3.1 Fast

Checks large-scale traffic logic, intersection pulses, blue-hour lighting, and whether the city avoids a miniature-model look.

Veo 3.1 Fast wins because it treats the city as a live simulation with vehicle-level decisions instead of a mostly animated plate.

Reference frame

Downtown traffic vs Veo 3.1 Fast frame

Original prompt

Prompt text is kept as the original benchmark source text.

Ultra cinematic aerial view of a modern Western metropolis at dusk.

The camera looks down over dense downtown streets filled with heavy traffic.

Hundreds of cars move continuously in multiple directions, headlights and taillights forming flowing light streams through the city.

Traffic moves fast but naturally - vehicles accelerate, slow down, merge, and change lanes realistically.

Intersections pulse with motion as cars pass through in waves.

Skyscrapers glow with warm window lights while neon signs flicker between buildings.

Golden sunset fades into blue hour as city illumination gradually takes over.

Soft atmospheric haze hangs between towers.

Drone-style wide 16:9 perspective.

Strong sense of urban life:
cars never stop,
streets always active,
light constantly shifting.

Cinematic color grading, high contrast, realistic lighting.

Dynamic motion everywhere:
moving vehicles,
scrolling traffic,
subtle camera drift.

Premium commercial look.

Hyper realistic.

Designed as first frame for fast-paced city video.

Side A

Seedance 2 Fast

Total score

68 / 100

41 / 60

Duration

10.07s

FPS

30

Resolution

1112x836

Side B

Veo 3.1 Fast

Total score

88 / 100

53 / 60

Duration

8.00s

FPS

24

Resolution

1280x720

Normalized score table

MetricSeedance 2 FastVeo 3.1 Fast
Prompt fit79
Motion / action79
Physics78
Camera89
Stability59
Cinematic finish89

Seedance 2 Fast readout

Seedance 2 Fast delivers strong color and framing, but the traffic mostly reads as simple layer motion rather than true urban flow.

Winning readout

Veo builds much richer lane changes, intersection pulses, and lighting shifts, which gives the whole city stronger physical credibility.

Fairness note

Veo runs longer in a harder city-simulation scenario, yet it keeps structures and traffic logic intact, so the runtime gap does not change the result.