VEO 3.1 AI 影片生成

VEO 3.1 is Google DeepMind's third-generation text-to-video model and the production successor to VEO 2. Unlike Sora's variable-length clips, every VEO 3.1 generation is a fixed 8 seconds — the model is optimized for that exact window, which makes it a strong fit for social formats and product loops.

All three variants — Lite, Fast and Quality — share the same core: 8-second output in 16:9 or 9:16, synchronized sound, and three input modes (text-to-video, image-to-video with a single first frame, and first/last frame control where you supply both endpoints and the model interpolates the motion between them).

The three tiers trade quality for cost, not resolution: Fast and Quality render 720p, 1080p or 4K and Lite renders 720p or 1080p, each at the same price whichever resolution you pick. VEO 3.1 Lite is the gateway at 24 credits per clip with sound always included — useful for storyboarding, mood-board iteration and animatic-style work. On Fast and Quality, sound is a switch that is on by default, because Google bills sound on top for these two tiers: VEO 3.1 Fast is 68 credits per clip with sound or 45 without, and VEO 3.1 Quality is 180 with sound or 90 without. Fast is the everyday workhorse; Quality is the slower, higher-fidelity tier for hero shots and final deliverables.

VEO is particularly strong at outdoor scenes, natural light, water and wind, and on-camera dialogue with lip-sync that holds up. It is weaker than Sora 2 Pro for indoor cinematic lighting and weaker than Seedance for stylized animation. First/last frame mode is unique within the VEO family on FlyAIgh — it gives you a guaranteed start and end pose, which is invaluable for tight editorial cuts.

On FlyAIgh, all three VEO 3.1 variants are available on every paid plan without a separate Google Cloud account, API project or quota request.