3 Ways to Access Google Veo 2

2. Prompt: An old man sitting alone at a train station as seasons change around him, time-lapse style, melancholic tone.

3. Prompt: A giant koi fish flying in the sky above a quiet village, clouds parting as it swims through the air

4. Prompt: Form the word “NICOLE” using bright-colored animal pool floaties in a swimming pool, overhead shot.

Prompt: “Form the word “NICOLE” using bright colored animal pool floaties in a swimming pool, overhead shot.” pic.twitter.com/Cg4ZDjUKjE

— Google Gemini App (@GeminiApp) April 15, 2025

Hands-on: Generate Cinematic Videos with Google Veo 2 on Vertex AI

While Google Veo 2 is accessible via platforms like Gemini, Whisk Animate, and VideoFX, developers can go a step further and integrate Google Veo 2 directly into their own applications using the Google GenAI SDK for Python via Vertex AI.

This hands-on tutorial walks you through building your own prompt-to-video generator in a Python environment like Google Colab or Jupyter Notebook.

Prerequisites

Before starting, ensure:

• You have a Google Cloud Project with Vertex AI API enabled
• You’ve set up billing and Cloud Storage
• Your environment is authenticated to access Google Cloud (e.g., via Google Colab or local gcloud auth)

Step 1: Install the Required Libraries

Install the GenAI SDK and a few helper libraries like mediapy for displaying videos in notebooks.

%pip install --upgrade --quiet google-genai
%pip install -q mediapy

Step 2: Authenticate (Colab-only)

If you’re running this in Google Colab, authenticate your Google account:

import sys
if "google.colab" in sys.modules:
    from google.colab import auth
    auth.authenticate_user()

Step 3: Import Python Libraries

Import everything you need to interact with Google Veo 2 and visualize the output.

import os
import time
import urllib
import matplotlib.pyplot as plt
import mediapy as media
from PIL import Image as PIL_Image
from google import genai

Step 4: Set Up Your Project and Client

You’ll need to connect to your Google Cloud project and specify the region.

PROJECT_ID = "[your-project-id]"  # Replace with your actual Project ID
if not PROJECT_ID or PROJECT_ID == "[your-project-id]":
    PROJECT_ID = str(os.environ.get("GOOGLE_CLOUD_PROJECT"))
LOCATION = os.environ.get("GOOGLE_CLOUD_REGION", "us-central1")
client = genai.Client(vertexai=True, project=PROJECT_ID, location=LOCATION)

Step 5: Define Helper Functions

These functions help download the generated video from your Cloud Storage and display it in the notebook.

def show_video(gcs_uri):
    file_name = gcs_uri.split("/")[-1]
    !gsutil cp {gcs_uri} {file_name}
    media.show_video(media.read_video(file_name), height=500)
def display_images(image):
    fig, axis = plt.subplots(1, 1, figsize=(12, 6))
    axis.imshow(image)
    axis.set_title("Starting Image")
    axis.axis("off")
    plt.show()

Step 6: Load Google Veo 2 Model

Use the specific model name required for video generation.

video_model = "veo-2.0-generate-001"

Step 7: Generate a Video from a Text Prompt

Now let’s generate a video using your own prompt. Here’s how:

prompt = "A dreamy, slow-motion shot of a Bengal tiger walking through misty jungle under golden morning light, as leaves fall and birds fly away.
"  # Example prompt
aspect_ratio = "16:9"            # "16:9" or "9:16"
output_gcs = "gs://your-bucket-name/cat-reading.mp4"  # Replace with your GCS path
operation = client.models.generate_videos(
    model=video_model,
    prompt=prompt,
    config=types.GenerateVideosConfig(
        aspect_ratio=aspect_ratio,
        output_gcs_uri=output_gcs,
        number_of_videos=1,
        duration_seconds=5,             # 5 to 8 seconds
        person_generation="dont_allow", # For safe content generation
        enhance_prompt=True             # Improves prompt quality
    ),
)

Monitor the operation until completion:

while not operation.done:
    time.sleep(15)
    operation = client.operations.get(operation)
    print(operation)
Finally, display the generated video:
if operation.response:
    show_video(operation.result.generated_videos[0].video.uri)

Output

This integration allows for scalable AI-driven video generation across custom apps, creative tools, or educational content. With Google’s SynthID watermarking and model enhancements, you also get ethical, production-ready media outputs.

While Google’s Google Veo 2 offers powerful video generation capabilities, several other platforms are pushing the boundaries of AI-generated video in unique ways. Here are some of the top alternatives worth exploring:

OpenAI’s Sora

Sora is OpenAI’s flagship text-to-video model that creates high-quality video clips up to 20 seconds long at 1080p resolution. It supports multimodal inputs text, images, and video and offers strong creative control through style presets and remix capabilities. Integrated directly into ChatGPT, it’s accessible to Plus and Pro users, making it a powerful option for creators looking for high fidelity and ease of use.

Runway Gen-3 Alpha

Runway’s Gen-3 Alpha is built for professionals in film and media. It features advanced text-to-video and image-to-video generation, along with photorealistic rendering and precise key-framing tools. With its ability to capture nuanced facial expressions and motion, it’s ideal for storytelling, content marketing, and cinematic experimentation.

Luma Labs’ Dream Machine

Dream Machine by Luma Labs stands out for its cinematic quality and natural motion generation. It transforms prompts or still images into dynamic video scenes with smooth camera work and lifelike animation. The interface is user-friendly, making it accessible for both beginners and experienced designers.

Kling AI

Developed by Kuaishou, Kling AI has quickly become known for its ability to produce HD videos up to two minutes long. It emphasizes realism and imagination, offering global access through a simple beta sign-up. Its visual quality and length support make it ideal for storytelling or creative content.

Hailuo AI

Hailuo AI is a free and lightweight video generator that produces short, imaginative clips from text prompts. It’s especially strong in prompt adherence and creativity, making it a great tool for users looking to experiment without cost or technical overhead.

These Google Veo 2 alternatives offer distinct strengths, from longer durations and high resolutions to professional-grade tools and easy accessibility. Whether you’re a filmmaker, educator, marketer, or hobbyist, exploring these platforms can open up exciting possibilities for AI-generated video storytelling.

How Google Veo 2 Compares to OpenAI’s Sora?

Google and OpenAI are now both in the generative video space. 

We put them to the test with this fun prompt:

Prompt: A tiny penguin barista serves frothy cappuccinos with latte art at an ice café, expertly balancing mugs on a tray as curious polar bears wait in line.

Google Veo 2 nailed the vibe, bringing the prompt to life with cinematic charm and stunning detail. The penguin waddles naturally, steams cappuccinos, and the polar bears react believably, all framed with soft snow and dynamic lighting. Sora, however, fell short. The penguin was stiff, the tray awkward, and the cappuccinos had candles. With flat visuals and no real storytelling, it simply missed the mark compared to Google Veo 2.

Here’s how Google Veo 2 stacks up against OpenAI’s Sora:

Google Veo 2 scored higher than Sora in over 58% of prompt alignment and preference comparisons, especially for scenes involving physical realism, human motion, or storytelling.

Tips for Crafting Better Prompts

To get the most out of Google Veo 2, keep these key tips in mind when writing your prompts:

• Be Descriptive: Mention colors, motion, time of day, environment, and other sensory details to create a vivid mental image.
• Use Cinematic Language: Include camera angles, transitions, or lighting styles (e.g., “wide-angle shot,” “overhead drone view,” “sunset backlight”).
• Set an Emotional Tone: Words like “tense,” “peaceful,” or “thrilling” help shape the mood and pacing of the video.
• Add Specific Actions: Don’t just describe the setting, include what the characters or objects are doing to bring the scene to life.

Example:

• Less Effective:  “A tiger in the jungle.”
• More Effective:  “A slow-motion tracking shot of a Bengal tiger prowling through dense, misty jungle foliage at dawn, sunlight filtering through the trees and glinting off its golden-orange fur as birds scatter in the background.”

Embedded Watermark for Transparency: SynthID in Google Veo 2

Google has integrated SynthID, a digital watermarking system, into all videos generated by Google Veo 2. This invisible marker is designed to identify content as AI-generated, even after it has been edited. The inclusion of SynthID is crucial as it promotes transparency, making it easier for platforms to identify and filter out manipulated media. Additionally, it supports the ethical use of AI by ensuring that AI-generated content can be traced. Even with common edits, such as cropping or applying filters, the watermark remains detectable, allowing for the identification of repurposed videos.

Also read: SynthID: Google is Expanding Ways to Protect AI Misinformation

Conclusion

Google’s Google Veo 2 is an awesome generative AI tool that enables the creation of cinematic-quality videos from detailed prompts. With features like realistic motion, cinematic lighting, and high prompt accuracy, it’s set to transform video production. The integration of SynthID ensures ethical AI use by embedding a digital watermark, making it easy to identify and filter AI-generated content. As Google Veo 2 expands across platforms, it promises to redefine how we create and engage with video content.

Frequently Asked Questions

Hi, I am Janvi, a passionate data science enthusiast currently working at Analytics Vidhya. My journey into the world of data began with a deep curiosity about how we can extract meaningful insights from complex datasets.