Veo 3.1 Fast

Launched October 14, 2025 alongside Veo 3.1, Veo 3.1 Fast is Google DeepMind's speed-optimized text/image-to-video model. It generates 8-second clips at 720p or 1080p, 24fps, with natively generated audio — dialogue with ~120ms lip-sync accuracy, sound effects, and ambient soundscapes at 48kHz — the first practical video model to do synchronized audio directly from a prompt. Reference-image support ('Ingredients to Video'), scene extension, and first/last-frame transitions give it filmmaker-grade control, and it plugs into the Gemini app, Flow, Vertex AI, and YouTube Shorts.\n\nLimitations: 8 seconds per generation (longer only via extension chains); Fast trades some fidelity for speed versus Veo 3.1 Standard. Per-second pricing adds up fast for long-form work, and like all video models it can struggle with complex physics and multi-character consistency over extended scenes.

Community-tested: leading lip-sync accuracy and native audio quality at launch