Veo 3.1 Lite Lands on Vertex AI at Half the Video Cost

If your team has been generating product demo clips or synthetic training footage with Veo 3.1 Fast, Google Cloud just gave you a reason to check the bill. Veo 3.1 Lite is now in public preview on Vertex AI, and it prices out at roughly $0.05 per second versus Fast's $0.10 per second, which is exactly the half-price cut the model's name promises. Google shipped it alongside a separate, standalone Veo upscaling capability for pulling existing footage up to 1080p or 4K.

Google Cloud's cheapest video model yet

Veo 3.1 Lite joins Veo 3.1 and Veo 3.1 Fast as the third tier in the family, and all three now generate native audio alongside video instead of needing a separate voice or sound pass. Google positions Lite specifically for "high-volume video applications": think batch-generating hundreds of ad variants, populating a catalog with product videos, or any workflow where you're calling the API thousands of times a month rather than producing a handful of hero clips. Veo 3.1 stays the pick for final production cuts, Fast sits in the middle for everyday quality at reasonable latency, and Lite trades some fidelity for volume. You can read Google's framing in the Vertex AI blog post.

Why the price cut matters for whoever owns the video feature

Vertex AI video generation pricing is per second of output, which means cost scales linearly with usage in a way that's easy to model and easy to blow through. At $0.40 per second, Veo 3.1 (standard) turns an 8-second clip into roughly $3.20. Fast brings that down to about $0.80. Lite brings it to about $0.40, cheap enough that generating video becomes a routine backend step instead of a special, gated feature you ration. If you're the one who has to defend the AI line item in a cost review, that's the number that makes the conversation easier.

Getting a Veo 3.1 Lite video into your pipeline

The model ID is veo-3.1-lite-generate-001, and it's callable through the same generate_videos method you'd already use for Veo 3.1 or Fast, just with a different model string:

from google import genai
from google.genai.types import GenerateVideosConfig

client = genai.Client(
    vertexai=True,
    project="<your-project-id>",
    location="us-central1",
)

operation = client.models.generate_videos(
    model="veo-3.1-lite-generate-001",
    prompt="A drone shot flying over a solar farm at sunrise",
    config=GenerateVideosConfig(
        aspect_ratio="16:9",
        output_gcs_uri="gs://<your-bucket>/veo-output/",
    ),
)

Video generation runs as a long-running operation, so poll operation.done and read the result from your Cloud Storage bucket once it finishes rather than expecting a synchronous response. Swapping models later, say from Lite to Fast for a premium tier of the same feature, is a one-line change to the model argument.

The upscaling half of the announcement

The new Veo upscaling capability is a separate, standalone tool: point it at existing footage and it pushes the resolution up to 1080p or 4K. Google says it works on Veo-generated clips, video from other AI models, and traditional camera footage. That's useful if your pipeline generates at Lite quality for volume and speed, then upscales only the handful of clips that get promoted to a customer-facing surface, rather than paying Fast or Standard rates for everything up front.

Gotchas before you flip the switch in production

Lite is in public preview, which on Google Cloud means it ships without the SLA guarantees and change-management discipline of a GA feature. Treat it the way you'd treat any preview API: pin the model version string, don't route 100% of production traffic to it without a fallback, and watch the release notes for breaking changes before you build deep dependencies on it. The upscaling capability is newer still, so confirm current availability before promising it to a customer. And because pricing is per second with no bundled minimum, run a quick cost projection against your expected monthly volume before you swap a whole pipeline over. Half price on paper is easy to undo with double the retries.

If you're running a video feature that has to scale, Veo 3.1 Lite is the tier that makes per-second billing at high volume finally look sane, and it's worth testing against your actual workload this week while it's still fresh in preview.