Veo 3.1 vs Sora 2: Same Prompt, Six Takes

Every “which is better” article you’ve read was written by someone who ran one prompt and picked a winner. We ran six identical prompts through both models — and the answer changes per round. Here’s the honest scorecard.

Round 1: Dialogue scene — Veo wins

Prompt: “A barista tells a regular their usual order is gone, rainy window behind them, she whispers the news.” Veo’s audio sync lip-locks the whisper; Sora 2’s dialogue drifted to ambient mumble. When words matter, Veo is the cast.

Round 2: Physics — Sora wins

Prompt: “A skateboarder lands a kickflip down five stairs, slow motion.” Sora 2 nailed board-body separation and landing impact. Veo’s board merged with the feet twice. Physics is still OpenAI’s home turf.

Round 3: Product hero — Veo, barely

Prompt: “Matte-black bottle rotating on wet stone, studio rim light, slow motion.” Both gorgeous. Veo’s rim light read more “shot on Alexa”; Sora’s texture was slightly more detailed. Toss-up — pick on taste.

Round 4: Crowd scene — Sora

Prompt: “Rush-hour crosswalk at dusk, hundreds of pedestrians, crane down.” Sora held individual coherence deeper into the frame. Veo’s background crowd dissolved earlier.

Round 5: Emotional close-up — Veo

Prompt: “Grandmother reads to her grandson, warm lamp light, gentle push-in.” Veo’s micro-expressions and lamp warmth were noticeably more filmic. This is the take that made our team gasp.

Round 6: Abstract art — Sora

Prompt: “Ink dispersing in water forms a phoenix, macro.” Sora’s morphing was more inventive; Veo stayed more literal.

Score: 3–3, and that’s the point. The models are converging on quality but diverging on specialty. Draft on Hailuo 02, run finals on both (/), keep the take that wins your scene. The comparison costs less than a sandwich — and it’s the only benchmark that matters: yours.