Veo 3 API Guide: Features, Pricing & Code (2026)

Google's Veo 3 has become one of the most talked about video models on the market, mostly because it does something almost no one else does well: it generates audio right alongside the video. Dialogue, sound effects, ambient noise, all synced to what's happening on screen, in the same generation call. If you're a developer trying to decide whether to build on it, this guide walks through what Veo 3 actually does, what it costs, how to get access, and how to call it from code.

What Is Veo 3

Veo 3 is Google's flagship AI video generation model, available through the Gemini API and Vertex AI. It takes a text prompt (or a text prompt plus a reference image) and produces a short video clip, generally up to 8 seconds, with resolution up to 1080p.

What sets it apart from most of the field is native audio generation. Models like Kling or early Seedance releases give you silent video that you then need to score, voice, and sound-design separately. Veo 3 generates the audio track as part of the same request, so a prompt describing a street musician playing violin in the rain can come back with the violin actually playing, rain hitting pavement, and ambient city noise, all synced to the visuals.

Google has positioned Veo 3 against Sora 2, Kling 3.0, and Seedance as the "premium" tier of video generation, and the audio feature is the main reason it holds that position. It's not the cheapest model, and it's not the fastest, but for anything where sound matters (dialogue-driven shorts, product demos with voiceover, ads) it's usually the first one developers reach for.

Key Features of the Veo 3 API

Native Audio Generation

This is the headline feature. Veo 3 can generate synced dialogue, sound effects, and ambient audio in the same pass as the video, rather than requiring a separate text-to-speech or sound design step. You can also turn audio off if you only need silent footage (useful if you're planning to add your own voiceover or music later, since skipping audio roughly halves the cost).

Resolution, Duration and Aspect Ratio

Veo 3 supports 720p and 1080p output, with duration options of 4, 6, or 8 seconds per clip. Aspect ratio is configurable so you can target landscape, square, or vertical formats depending on where the video is headed.

Prompt Adherence and Camera Control

Veo 3 is known for following detailed prompts closely, including camera direction (slow pan, tracking shot, close-up) and cinematic language. It also supports negative prompts, so you can explicitly tell it what to avoid in the frame.

Image-to-Video Mode

Instead of generating purely from text, you can pass a reference image as the starting frame and have Veo 3 animate from there. This is useful for product shots, brand assets, or any case where you need the first frame to match something specific.

Veo 3 API Pricing

Official Pricing via Gemini API and Vertex AI

Google bills Veo 3 per second of output, and the rate depends on the tier and whether audio is included. Recent public pricing puts the Lite tier around $0.05 per second at 720p, the Fast tier around $0.10 per second, and the Standard tier around $0.40 per second for video-only output at 720p or 1080p. Adding native audio increases the cost by roughly 50 percent, which puts an 8-second Standard clip with audio at around $6. Because Google adjusts these tiers periodically as new Veo versions ship, it's worth checking the current Gemini API and Vertex AI pricing pages before budgeting a production workload.

Fast vs. Standard Tier

The Fast tier trades some quality and prompt adherence for a lower per-second cost and quicker turnaround, which makes sense for prototyping or high-volume, lower-stakes content. Standard is the one most teams reach for once they're shipping to production, particularly when audio quality matters.

Hidden Costs to Watch

A few things catch teams off guard: retries on failed or low-quality generations (you're often billed for the attempt even if you discard the result), storage and egress if you're pulling large volumes of video out of cloud storage, and the cost difference between silent and audio output, which is easy to forget when estimating a budget.

How to Access the Veo 3 API

Official Google Routes

Veo 3 is available directly through the Gemini API and through Vertex AI, Google Cloud's ML platform. Both require a Google Cloud account with billing enabled, and Vertex AI in particular is built around quota and region settings, so it's worth checking availability in your target region before committing to it as your only route.

Aggregator APIs

If you don't want to manage a direct Google Cloud integration, or you want the option to compare Veo 3 output against other video models without building a second integration, aggregator APIs are the other route. Apiframe, for example, exposes Veo 3 through a single unified video endpoint alongside Kling, Sora 2, Seedance, Runway, and others, so you can switch models by changing one parameter in your request instead of maintaining separate SDKs and auth flows per provider.

Which Route Fits Your Use Case

If you're already deep in the Google Cloud ecosystem and only ever plan to use Veo 3, going direct through Vertex AI keeps things simple. If you want to A/B test Veo 3 against other models, need one bill instead of several, or don't want to deal with waitlists and quota requests, an aggregator is usually less friction to get started with.

Veo 3 API Code Example

Here's what a Veo 3 generation request looks like through Apiframe's video API, which wraps Veo 3 behind a single POST /v2/videos/generate endpoint alongside every other supported video model:

import requests

response = requests.post(
    "https://api.apiframe.ai/v2/videos/generate",
    headers={
        "X-API-Key": "afk_your_api_key_here",
        "Content-Type": "application/json",
    },
    json={
        "prompt": "a street musician playing violin in a rainy alley, cinematic",
        "model": "veo-3",
        "veoParams": {
            "duration": 8,
            "generate_audio": True,
            "resolution": "1080p",
        },
    },
)
data = response.json()
print(data)  # {"jobId": "...", "status": "QUEUED"}

The request returns immediately with a jobId and a QUEUED status. Video generation is asynchronous, so you either poll the job or supply a webhookUrl and let the API notify you when it's done. Polling looks like this:

import time

job_id = data["jobId"]

while True:
    job = requests.get(
        f"https://api.apiframe.ai/v2/jobs/{job_id}",
        headers={"X-API-Key": "afk_your_api_key_here"},
    ).json()

    if job["status"] == "COMPLETED":
        print(job["result"]["videoUrl"])
        break
    elif job["status"] == "FAILED":
        print(job["error"])
        break

    time.sleep(3)

Once the job completes, result.videoUrl points to a CDN-hosted MP4. A few parameters worth knowing about on the veoParams object: duration accepts 4, 6, or 8 seconds, resolution is "720p" or "1080p", generate_audio defaults to true but can be set to false for silent output, and image lets you pass a reference image URL for image-to-video generation. If you'd rather not poll, adding a webhookUrl and webhookEvents: ["completed", "failed"] to the request body triggers a callback to your server instead.

Veo 3 vs. Other Video Models

Pricing across the industry shifts often as providers adjust tiers, so treat these as rough positioning rather than numbers to build a budget on. The practical takeaway: Veo 3 sits toward the premium end of the market, and its main differentiator is audio quality, not being the cheapest or fastest option. If price per second is the deciding factor, Kling or Seedance tend to come in lower. If native, well-synced audio is the deciding factor, Veo 3 is usually the first one worth testing.

Use Cases for the Veo 3 API

Marketing teams use Veo 3 for short-form ads and social content where a voiceover or synced sound effect adds real production value without a separate audio pass. Product teams use it for demo videos where a narrated walkthrough needs to match on-screen action. It also shows up a lot in storyboarding and previz work, where directors want a quick sense of how a scene will look and sound before committing to a full shoot. And because dialogue generation is native, it's a natural fit for short-form narrative content that leans on a character actually speaking, rather than silent b-roll with text overlays.