Cafe, three shots
The prompt “Three consecutive shots in the same cafe: wide of the interior, medium of a woman in a red coat at a table, close-up of her hands around a cup. Same woman, same cafe, same lighting.”
Character and location must survive across cuts. This is the capability that separates 2026 models from 2025 ones.
- Models tested
- 16
- Top score
- 5.00 13 models tied
- Average
- 4.97
- Lowest
- 4.70 Runway Gen-4.5
Every model, same prompt
The clips on screen play together from the first frame, muted. Open one for sound, controls and our notes.
Kling 4.0
KuaishouThree shots in the asked order, wide, medium, then hands around the cup, with cuts at about 2.9 and 5.8 seconds and the same woman, red coat and cafe. Checked frame by frame.
Wan 3.0
AlibabaThree shots in the asked order, wide, medium, then hands around the cup, with cuts at about 2.5 and 4.5 seconds and the same woman, red coat and cafe. Checked frame by frame.
Vivix W1
VivixThree shots in the asked order, wide, medium, then hands around the cup, with the same woman, the same red coat and the same cafe. Checked frame by frame.
FLUX 3
Black Forest LabsThree shots in the asked order, wide, medium, then hands on the cup, with the same woman in the red coat and the same cafe throughout. Checked frame by frame.
Kling 3.0 Turbo
Kuaishou
MiniMax H3
MiniMaxCoat, mug and table stay identical across all three shots.
MiniMax Hailuo 02
MiniMaxCoat, table and cafe interior match seamlessly across all three shots.
Google Veo 3.1
GoogleCoat, cup and table grain match exactly across the cut.
Seedance 2.5
ByteDanceSame woman, coat and mug hold identical across all three shots.
Seedance 2.0
ByteDanceSame woman, coat, mug and table hold identical across all three shots.
Grok Video
xAI3 shots via cross-dissolve, woman/coat/mug identical throughout.
Kling 3.0
Kuaishou3 real cuts, woman/coat/cup/cafe identical throughout.
OpenAI Sora 2
OpenAIWoman, coat, cafe and latte art identical across all 3 shots.
Google Gemini Omni
GoogleReal cuts (wide -> medium -> close-up) with fully consistent identity, coat, mug and cafe across all three -- the intended multi-shot execution.
PixVerse v5.5
PixVerseReal cut used between the wide and close-up shots (unlike most other models on this prompt, which render one continuous take); identity/setting fully consistent across the cut.
Runway Gen-4.5
RunwayRendered as one continuous push-in rather than distinct cuts; otherwise clean.
Ranked by our blind vision-model judge; models with the same score share a position. Read how we test.