Get the verdict
Same-prompt test 6 · video · October 2026

Milk pouring into black coffee, macro: which AI does it best?

Runway Gen-4.5 wins with 8.3/10. Behind it: Kling 3.0 (8.0), PixVerse V6 (8.0), Veo 3.1 Fast (6.5). One 4-second clip per model, unedited, with whatever audio the model made.

The promptA hand pouring milk into black coffee, the swirl in macro, top-down.
#ModelOveralladherencemotionqualityartifactsCostRender
1Runway Gen-4.58.38898$0.4868s
2Kling 3.08.08888$0.52130s
3PixVerse V68.08888$0.2146s
4Veo 3.1 Fast6.57676$0.6068s

Best per category: adherence Runway & Kling & PixVerse (8) · motion Runway & Kling & PixVerse (8) · quality Runway (9) · artifacts Runway & Kling & PixVerse (8)

Every output, ranked

#1 Runway Gen-4.5

Runway Gen-4.5 8.3
4s · ~48 cr ($0.48) · 68s render

adherence 8motion 8quality 9artifacts 8
Beautiful swirl in a mug; angled rather than top-down.

#2 Kling 3.0

Kling 3.0 8.0
4s · audio · ~52 cr ($0.52) · 130s render

adherence 8motion 8quality 8artifacts 8
Near top-down glass, clean realistic swirl.

#3 PixVerse V6

PixVerse V6 8.0
4s · audio · 48 PixVerse cr ($0.21) · 46s render

adherence 8motion 8quality 8artifacts 8
Realistic milk swirl in black coffee, near top-down; pitcher partly in frame.

#4 Veo 3.1 Fast

Veo 3.1 Fast 6.5
4s · audio · ~60 cr ($0.60) · 68s render

adherence 7motion 6quality 7artifacts 6
Milk forms unrealistic blobs.

Try Runway ↗

⚖️
Outputs are shown exactly as the API returned them (4s, 720p, default settings). Scores are editorial judgments on a fixed rubric, given without knowing which model made each output. One run is a snapshot, not a guarantee. Methodology

FAQ

Which AI is best for "AI video of liquid pouring"?

In our October 2026 test of this exact prompt, Runway Gen-4.5 wins with 8.3/10, ahead of Kling 3.0 (8.0), PixVerse V6 (8.0), Veo 3.1 Fast (6.5). Runway Gen-4.5's output: Beautiful swirl in a mug; angled rather than top-down.

What exact prompt was used?

“A hand pouring milk into black coffee, the swirl in macro, top-down.” Every model got the same prompt with default settings, 4 seconds at 720p, and one best-effort retry only if a generation failed.

How were the outputs scored?

Each output was scored without knowing which model made it, from 1 to 10 on adherence, motion, quality, artifacts. The overall score is the average of those.