Curiosity · AI Model
Google Veo 2
Veo 2 is Google DeepMind's second-generation text-to-video model. It generates cinematic clips of up to 8 seconds at 4K, with camera-motion control, cinematography keywords, and high physical realism. Veo 2 is served through Google Cloud Vertex AI, the Gemini app, and VideoFX in Google Labs.
Model specs
- Vendor
- Family
- Veo
- Released
- 2024-12
- Context window
- 1 tokens
- Modalities
- text, vision, video
- Input price
- n/a
- Output price
- n/a
- Pricing as of
- 2026-04-20
Strengths
- 4K output with strong cinematic lighting
- Explicit camera-motion keywords (dolly, crane, handheld)
- Tight integration with Vertex AI, Gemini app, and Google Labs
- SynthID watermarking for provenance
Limitations
- No token context — priced by seconds and resolution
- Clips capped at 8 s — longer pieces need stitching
- Audio is generated separately (Veo 3 adds native audio)
- Enterprise availability varies by region
Use cases
- Cinematic ad concepts and brand films
- B-roll for explainer videos and documentaries
- Storyboarding with cinematography keywords
- Vertex-native enterprise video pipelines
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| Video Arena ELO (2025) | ≈1200 | 2025 |
| Max duration × resolution | 8 s × 4K | 2025 |
Frequently asked questions
What is Veo 2?
Veo 2 is Google DeepMind's second-generation text-to-video model. It generates cinematic clips of up to 8 seconds at 4K, with camera-motion controls and high physical realism, available through Vertex AI, the Gemini app, and Google Labs VideoFX.
How long can Veo 2 clips be?
Up to 8 seconds per generation. Longer videos are built by stitching multiple generations, and Veo 3 extends duration and adds native audio for longer productions.
Does Veo 2 support camera control?
Yes — prompts accept camera-motion keywords (dolly-in, crane, handheld, locked-off) and shot-type directives (close-up, wide, over-the-shoulder), giving directors explicit handles on cinematography.
How is Veo 2 priced?
Via Vertex AI, Veo 2 is billed per second of generated video, with tier-based pricing for 1080p vs 4K and different durations. Consult the Vertex AI pricing page for current rates.
Sources
- Google DeepMind — Veo 2 — accessed 2026-04-20
- Vertex AI — Veo generation docs — accessed 2026-04-20