Google has released Veo 3.1, a subtle yet impactful update to its AI video generation model. While not a complete overhaul, the new version brings meaningful improvements in realism, audio quality, facial expressions, and creative control—making it a compelling choice for creators and developers. Veo 3.1 introduces enhanced visual fidelity, particularly in simulating physics. Fluid dynamics, such as water splashes and ripples, now appear far more natural and consistent. For example, a prompt describing a woman performing a backflip into a pool results in a video where the water interaction is highly realistic, with proper splash patterns, reflections, and smooth motion. The model also handles complex body movements accurately, avoiding distortions or unnatural limb positions. Audio is now better synchronized with on-screen action, adding a layer of immersion that was missing in earlier versions. Another standout feature is the improved handling of facial expressions. Veo 3.1 generates more dynamic and emotionally expressive faces, especially in close-up shots. A side-by-side comparison with the previous model shows a clear difference—faces now convey a wider range of emotions with greater nuance. In one test, a woman transitions from neutral to frowning, screaming, and laughing hysterically, with realistic eye movements, mouth shapes, and facial muscle engagement. This level of expressiveness is ideal for character-driven scenes, storytelling, or even animation reference. The most innovative addition is the “Ingredients to Video” feature. This allows users to upload reference images—such as a person, clothing, or object—and use them as visual anchors in a video. The model maintains consistency across multiple shots, ensuring the subject, outfit, or style remains unchanged. For instance, a prompt for a fashion photoshoot featuring an Asian man in a black puffer jacket and sunglasses produces a video where the jacket, sunglasses, and pose remain consistent throughout. The lighting, camera movement, and natural gestures all align with the description, creating a polished, professional result. Google also launched Veo 3.1 Fast, a lighter, faster version designed for quicker turnaround and lower cost, though with slightly reduced output quality. Both models are available through Google AI Studio, Vertex AI, Google Flow, and third-party platforms like Pollo AI, Fal AI, and Replicate. Currently, Pollo AI offers a 50% discount on Veo 3.1, making it an accessible entry point for experimentation. Users can access the model via text-to-video or image-to-video tools, with options to adjust length, aspect ratio, resolution, and audio settings. While the changes in Veo 3.1 are incremental, their combined effect is significant. The model now competes more directly with OpenAI’s Sora, especially in areas like physics simulation and emotional realism. With features like multi-frame reference and video extension, Veo 3.1 opens new possibilities for content creation, product showcases, and AI-driven storytelling. It’s not a revolution—but it’s a strong step forward.