xAI
Grok Video
Grok Video is xAI entry into video generation, bundled with the wider Grok product rather than sold standalone.
- Best category
- On-screen text 5.00
- Weakest category
- Hands 3.88
- Clips scored
- 40
- Generation time
- 85s
At a glance
Strengths
- Fastest generation of any model we tested, averaging 85 seconds per clip.
- 1 attempt on average per usable clip.
Watch-outs
- No native audio generation.
Silent clips where turnaround speed matters more than polish.
Capabilities
- Native audio
- No
- Multi-shot sequences
- No
- Image to video
- Yes
- Public API
- Yes
Benchmark scores
How we test →| Dimension | Score | |
|---|---|---|
| Overall | 4.72 / 5 | |
| Prompt fidelity | 4.70 | |
| Visual quality | 4.80 | |
| Physics & motion | 4.44 | |
| Subject consistency | 4.71 | |
| Audio & sync | 5.00 |
- Average attempts to a usable clip
- 1.00
- Average generation time
- 85s
- Scored runs
- 40
| Prompt category | Score | Runs | |
|---|---|---|---|
| On-screen text | 5.00 | 4 | |
| Lip-sync | 5.00 | 4 | |
| Camera | 5.00 | 4 | |
| Image to video | 5.00 | 4 | |
| Liquids | 4.83 | 4 | |
| Multi-shot | 4.71 | 4 | |
| Style | 4.67 | 4 | |
| Animation | 4.57 | 4 | |
| Faces | 4.54 | 4 | |
| Animals | 4.54 | 4 | |
| Hands | 3.88 | 4 |
Scores out of 5, judged blind. Bars start at 3.0 so the gaps are visible.
Cite this
Free to cite and reuse with a link back. Data is updated as new runs and prices come in, so include the date.
Genativity. (2026). Grok Video benchmark results. Retrieved October 23, 2026, from https://genativity.com/model/grok-video<a href="https://genativity.com/model/grok-video">Grok Video benchmark results</a> (Genativity)The clips
Every score above links to the actual generation it came from. Nothing here is a mockup.
Hands avg 3.88/5
Chopping a pepper
FAILURE MODE: anatomical hallucination, three separate hands appear manipulating the knife/pepper; slices duplicate and slide.
Counting fingers
Five digits throughout, but one finger rises first and the rest follow together rather than one by one.
Fretting a guitar chord
Plausible fret contact through a real chord change, but the clip cuts from the close-up to a wider shot halfway through.
Tying shoelaces
Correct anatomy, but the knot morphs together unnaturally rather than being threaded.
Faces avg 4.54/5
Face in a crowd
Sharp subject, coherent blurred crowd.
Elderly portrait, slow blink
Exceptionally stable, natural blink and wrinkle detail.
Glasses reflection
Identity coherent, but the reflection stays static instead of tracking the head turn.
Turn and smile
Identity holds through the turn/smile, but a glitching yellow artifact appears in one eye late in the clip.
Liquids avg 4.83/5
Pouring coffee
Accurate turbulence, foam and steam, no artifacts.
Honey drizzle
Correct high-viscosity laminar flow and pooling.
Ice melting in whisky
Good refraction/condensation; the cube shrinks a touch faster than real melting would.
Wave on rocks
Authentic curl, spray and foam drainage.
On-screen text avg 5/5
Book chapter heading
"CHAPTER ONE" exact and stable, succeeds where Kling failed on this exact prompt.
Neon sign OPEN 24H
"OPEN 24H" exact and stable.
Bakery sign
"FRESH BREAD DAILY" exact and stable.
Printed t-shirt
"SLOP" exact, deforms naturally with fabric motion.
Multi-shot avg 4.71/5
Cafe, three shots
3 shots via cross-dissolve, woman/coat/mug identical throughout.
Same bench, day and dusk
FAILURE MODE: single continuous shot, stays in midday light, the requested dusk transition never happens.
Watch, wide then macro
Watch dial, hands and date window identical from wide to macro.
Alley, two angles
2 shots, jacket/alley identical across the cut.
Lip-sync avg 5/5
Bearded man, long line
Exact line, sync holds through beard occlusion.
Direct address to camera
Exact line, phoneme-accurate sync.
Sung line
Sung line matches, articulation tracks the melody.
Two-person exchange
Correct turn-taking, both speakers synced across the cut.
Camera avg 5/5
Crane to skyline
Real reveal-style crane rise, stable architecture, succeeds where Sora failed on this exact prompt.
Dolly in to a door
Clean dolly-in, fully rigid geometry.
Handheld follow
Convincing handheld follow, stable terrain.
360 orbit
Genuinely completes the 360 with stable geometry, succeeds where Sora failed on this exact prompt.
Animals avg 4.54/5
Pigeon takeoff
Correct wing/leg anatomy; wingbeat rate a touch slow for a real pigeon.
Cat jumping to a counter
Full ballistic arc with real weight absorption on landing, the best result across all 3 models on this prompt.
Dog shaking off water
Realistic radial water-shedding and fur deformation; minor blur at peak shake speed.
Galloping horse
Correct anatomy and stride extension; hooves slide slightly on the ground.
Style avg 4.67/5
Anime walk cycle
Consistent cel-shaded character.
Claymation wave
Convincing clay texture and tool marks.
1970s film look
FAILURE MODE: defaults to clean modern digital look, the requested 1970s grain/halation barely appears.
Rotating mug
Rigid identity, clean studio look.
Image to video avg 5/5
Animate a landscape
Terrain static, clouds/grass move correctly.
Animate a vintage photo
Grain/tone preserved, no colourisation, correct 16:9.
Animate a portrait
Faithful to source, natural blink and hair motion.
Product photo to spin
Full 180 degrees, consistent product identity, correct 16:9 (no aspect-ratio drift here).
Animation avg 4.57/5
3D feature film look
Warm feature-film lighting, a head tilt, a raised palm and a smile with a soft face glow. The butterfly lands at about 5.5 seconds and then vanishes into the palm half a second later.
Anime sword draw
In a dojo she draws at 2.5 seconds, runs at the camera and ends in a lunge with hair and jacket flowing. The blade stays straight and rigid throughout; the final pose is held for about 3.5 seconds.
Squash and stretch
Strong squash on landing, a tall stretch on the rebound and a clear surprised face at the end. On the rebound the ball passes in front of the table it was sitting behind a moment earlier.
2D walk cycle
A steady, even cycle with the legs always in frame, but it is a bounding trot rather than a walk. The design never changes.
Where to use it
The same model costs different amounts depending on the platform you reach it through.
