Continue existing video clips with consistent motion, lighting, and identity using Google Veo 3-1.

• Routes to two endpoints: full-quality extend-video for hero shots and final delivery, or faster fast/extend-video for cost-controlled iteration and drafts
• Accepts a source video URL and a prompt describing what happens next; the model generates a continuation clip that maintains physics, framing, and subject consistency
• Supports chaining multiple extend calls shot-by-shot to build longer narratives, with identity holding reliably across 3–4 consecutive extends
• Prompts should anchor camera movement explicitly and describe only the next beat, not re-describe the existing scene; one main action per extend call yields best results