Google Gemini

Generate videos with Google's Omni Flash and Veo 3.1 models

Generate videos using Google’s Omni Flash and Veo 3.1 models through the Gemini API.

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="gemini/veo-3.1-generate-preview")
video = agent.start("A sunrise over mountains", output="sunrise.mp4")

Authentication

Set your Gemini API key before running.

export GOOGLE_API_KEY=your-key
# GEMINI_API_KEY also works

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="gemini/veo-3.1-generate-preview")
video = agent.start("A city skyline at night", output="skyline.mp4")

Models

Three Veo 3.1 tiers cover cinematic quality down to cheap iteration.

Which Veo 3.1 tier should I pick?

Deprecated

Migrate to the veo-3.1 IDs above. These older models are being retired.

Gemini Omni Flash

Gemini Omni Flash (gemini-omni-1.1-flash) is Google’s recommended default video model. Use Veo 3.1 for cinematic pipelines and interpolation.

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="gemini/gemini-omni-1.1-flash")
video = agent.start("A marble rolling on a chain-reaction track", output="marble.mp4")

Veo 3.1 features

Veo 3.1 adds frame interpolation and reference images. Both pass straight through generate() — no extra setup.

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="gemini/veo-3.1-generate-preview")

# First + last frame interpolation (Veo 3.1)
agent.generate(
    prompt="Sunrise timelapse",
    input_reference=start_image,  # first frame
    last_frame=end_image,         # last frame
)

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="gemini/veo-3.1-generate-preview")

# Up to 3 reference images
agent.generate(
    prompt="A cinematic dolly shot of the character",
    reference_images=[img1, img2, img3],
)

Pricing & billing

Gemini Veo bills per generated second in USD on the paid API. You are charged only when a generation succeeds.

Parameters

video = agent.generate(
    prompt="Your description"
)

Features