Area 01 · 12 jobs

Best AI for video, job by job

We don't rank models in general. We test them one job at a time: the same inputs and settings for every model, graded blind, with the cost of each usable result. 8 of these jobs are being tested now.

Testing now

Why it's hard

Faces, hands and skin look waxy; motion feels off

What we test

20 briefs: walking, talking, eating, close-ups

See the results so far →

Talking presenter (口播 / digital human)

Testing now
Why it's hard

Lip-sync drifts, gestures robotic, voice mismatched

What we test

60 s script in EN and 中文, same face and voice

Product ad, 9:16

Testing now
Why it's hard

Product shape and logo change; text on screen garbled

What we test

Same product photo, 10 s ad, on-screen text

Comic drama (漫剧)

Testing now
Why it's hard

Art style and characters drift between episodes

What we test

6-shot scene from a character sheet

Short drama (短剧), live action

Testing now
Why it's hard

Multi-shot continuity, acting, costume consistency

What we test

6-shot dialogue scene, 2 characters

Image-to-video from a reference

Testing now
Why it's hard

Model ignores the reference or changes identity

What we test

Our standard reference set (see section 4)

Same character across shots

Testing now
Why it's hard

The #1 creator complaint: the face changes every shot

What we test

5 shots, 1 reference sheet, identity scored

Lip-sync and dubbing into another language

Testing now
Why it's hard

Mouth doesn't match, voice loses emotion

What we test

30 s clip, EN↔中文↔JA

Coming in month 2

Real-estate walkthrough from renderings

Month 2
Why it's hard

Geometry warps, rooms morph

What we test

3 renderings to a 20 s fly-through

Music video and dance

Month 2
Why it's hard

Motion off-beat, limbs break

What we test

Same 15 s track, dance and performance

Action and physics

Month 2
Why it's hard

Fights, sports and cars break physics

What we test

10 motion briefs

Coming in month 3

Editing existing footage (VFX, relight, extend)

Month 3
Why it's hard

Edits leak outside the subject, flicker

What we test

Same 3 source clips

Models in this area

The field as of 29 Oct 2026. Prices are list prices.

ModelLatest versionList priceMainland China
Seedance 2.5ByteDance23 Jun 2026from $0.134/sYes
Kling 4.0Kuaishou15 Oct 2026WaveSpeed $0.157–0.235/sYes
Wan 3.0Alibaba24 Aug 2026$0.05–0.20/sYes
Gemini Omni 1.1 FlashGoogle~27 Aug 2026~$0.10/s at 720p?
MiniMax Hailuo H3 / H3 Max3 Aug 2026$0.05–0.08/s?
FLUX 3 Video / MaxBlack Forest LabsMax 20 Oct 2026??
HeyGen Video 1late Sep 2026~$0.015–0.03/s?
PixVerse V630 Mar 2026??

How we choose and grade: our methodology.