Mastering Veo 3.1 Video Extension API: Complete Guide to Generating 148-Second Long Videos with 7-Second Incremental Extensions

Want to use AI to generate long videos over 8 seconds, but finding that single generations have time limits? This is a common bottleneck for AI video creators. In this post, we'll deep dive into Google Veo 3.1's extend video capability, helping you master the complete technical solution for extending an 8-second short video into a 148-second long video via API.

Core Value: After reading this article, you'll learn how to use the Veo 3.1 extend API for video extension, understand the 7-second incremental extension mechanism, and be able to independently implement AI video generation up to 148 seconds long.

Core Points of the Veo 3.1 Video Extension API

Before we dive into the technical details, let's look at the key parameters and constraints of Veo 3.1's extend capability.

Detailed Explanation of Veo 3.1 Video Extension Principles

Veo 3.1's Scene Extension feature uses a clever continuous generation mechanism:

1. Last 1-second sampling: The system extracts visual features from the last second (24 frames) of the input video.
2. Continuity modeling: Predicts the next 7 seconds of video content based on these features.
3. Seamless stitching: Merges the newly generated 7-second video with the original video into one complete file.
4. Iterative accumulation: The output of each extension can serve as the input for the next extension.

This design ensures that the video maintains visual continuity even after multiple extensions, including:

• Smooth transition of character movements
• Natural transition of scene lighting
• Consistency of background elements
• Continuity of audio (if any)

🎯 Technical Tip: If you need to call the Veo 3.1 video extension API, you can get a unified interface through the APIYI platform, which supports convenient access to Google's suite of video generation models.

Veo 3.1 Extend API Technical Specifications

Input Constraints

Before calling the Veo 3.1 extend API, you've got to ensure your input video meets the following specifications:

Important Note: The Gemini API's extend feature only supports videos generated by Veo as input. If you try to use videos from other sources (like phone recordings or other AI-generated videos), the API will throw a validation error.

Output Specifications

Quick Start: Veo 3.1 Video Extension API

Setting Up Your Environment

Before we dive in, make sure you've got the following ready:

1. A Google AI Studio or Vertex AI account.
2. Access to the Gemini API.
3. Python 3.8+ installed on your machine.
4. The google-genai SDK installed.

pip install google-genai

Minimalist Code Example

Here's the most basic code you'll need to call the Veo 3.1 extend API:

from google import genai
import time

# 初始化客户端
client = genai.Client(
    api_key="YOUR_API_KEY"
    # 也可使用 APIYI apiyi.com 统一接口
)

# 步骤1: 先生成原始视频
print("正在生成原始视频...")
initial_operation = client.models.generate_videos(
    model="veo-3.1-generate-preview",
    prompt="一只金色的猎鹰在蓝天中翱翔,阳光穿透云层",
)

# 等待生成完成
while not initial_operation.done:
    time.sleep(30)
    initial_operation = client.operations.get(initial_operation)

initial_video = initial_operation.result.generated_videos[0]
print(f"原始视频生成完成,时长: 8秒")

# 步骤2: 扩展视频
print("正在扩展视频...")
extend_operation = client.models.generate_videos(
    model="veo-3.1-generate-preview",
    prompt="猎鹰继续在空中盘旋,俯冲捕捉猎物",
    video=initial_video
)

# 等待扩展完成
while not extend_operation.done:
    time.sleep(30)
    extend_operation = client.operations.get(extend_operation)

extended_video = extend_operation.result.generated_videos[0]
print(f"视频扩展完成,总时长: 15秒")

🚀 Quick Start: We recommend using the APIYI (apiyi.com) platform to quickly test the Veo 3.1 API. It provides an out-of-the-box interface, so you can get integrated without any complex configuration.

from google import genai
import time
import os

class Veo31VideoExtender:
    """Veo 3.1 视频扩展器 - 支持最长 148 秒视频生成"""

    def __init__(self, api_key: str):
        self.client = genai.Client(api_key=api_key)
        self.model = "veo-3.1-generate-preview"
        self.max_extensions = 20  # 最大扩展次数
        self.extension_duration = 7  # 每次扩展 7 秒

    def generate_initial_video(self, prompt: str, aspect_ratio: str = "16:9"):
        """生成初始视频"""
        print(f"[1/2] 生成初始视频...")
        print(f"    提示词: {prompt[:50]}...")

        operation = self.client.models.generate_videos(
            model=self.model,
            prompt=prompt,
            config={
                "aspect_ratio": aspect_ratio,
                "number_of_videos": 1
            }
        )

        video = self._wait_for_completion(operation)
        print(f"    初始视频生成完成 (8秒)")
        return video

    def extend_video(self, video, prompt: str, target_duration: int = 148):
        """
        扩展视频到目标时长

        Args:
            video: 输入视频对象
            prompt: 扩展提示词
            target_duration: 目标时长(秒),最大 148 秒

        Returns:
            扩展后的视频对象
        """
        # 计算需要的扩展次数
        initial_duration = 8
        needed_duration = target_duration - initial_duration
        extensions_needed = min(
            (needed_duration + self.extension_duration - 1) // self.extension_duration,
            self.max_extensions
        )

        print(f"[2/2] 开始视频扩展...")
        print(f"    目标时长: {target_duration}秒")
        print(f"    需要扩展: {extensions_needed}次")

        current_video = video
        current_duration = initial_duration

        for i in range(extensions_needed):
            print(f"    扩展进度: {i+1}/{extensions_needed}")

            operation = self.client.models.generate_videos(
                model=self.model,
                prompt=prompt,
                video=current_video
            )

            current_video = self._wait_for_completion(operation)
            current_duration += self.extension_duration

            print(f"    当前时长: {current_duration}秒")

        final_duration = min(current_duration, 148)
        print(f"视频扩展完成! 最终时长: {final_duration}秒")
        return current_video

    def _wait_for_completion(self, operation, check_interval: int = 30):
        """等待操作完成"""
        while not operation.done:
            time.sleep(check_interval)
            operation = self.client.operations.get(operation)

        if operation.result.generated_videos:
            return operation.result.generated_videos[0]
        raise Exception("视频生成失败")

    def download_video(self, video, output_path: str):
        """下载视频到本地"""
        print(f"下载视频到: {output_path}")

        # 获取视频内容
        video_data = self.client.files.download(video.video)

        with open(output_path, 'wb') as f:
            f.write(video_data)

        print(f"下载完成! 文件大小: {os.path.getsize(output_path) / 1024 / 1024:.2f} MB")


# 使用示例
if __name__ == "__main__":
    # 初始化扩展器
    extender = Veo31VideoExtender(api_key="YOUR_API_KEY")

    # 生成初始视频
    initial_video = extender.generate_initial_video(
        prompt="夕阳下的海边,金色的阳光洒在波光粼粼的海面上,一艘帆船缓缓驶向远方",
        aspect_ratio="16:9"
    )

    # 扩展到 60 秒
    extended_video = extender.extend_video(
        video=initial_video,
        prompt="帆船继续前行,天空逐渐变成橙红色,海鸥在船边盘旋",
        target_duration=60
    )

    # 下载视频
    extender.download_video(extended_video, "extended_video_60s.mp4")

Advanced Tips for the Veo 3.1 Video Extension API

Prompt Strategies: Ensuring Video Continuity

When you're extending a video, how you write your prompt directly impacts the result. Here are some best practices to keep things smooth:

Audio Processing Considerations

Veo 3.1 supports native audio generation, but there are a few limitations to keep in mind when using the extend feature:

1. Background Sound Effects: It's quite good at continuing ambient sounds and background music.
2. Dialogue/Vocals: If there's no vocal audio in the last second of the source video, the extension won't have it either.
3. Audio Coherence: The system tries its best to maintain a consistent audio style.

💡 Pro-tip: If your video needs continuous dialogue, make sure the last second of the original video contains speech; otherwise, the extended portion will likely only have background sound effects.

Batch Extension and Cost Optimization

If you need to generate a lot of long-form content, consider these optimization strategies:

# 批量扩展优化示例
def batch_extend_videos(video_list, prompts, target_duration=60):
    """
    批量扩展视频
    通过 APIYI apiyi.com 平台可以获得更优惠的批量调用价格
    """
    results = []

    for i, (video, prompt) in enumerate(zip(video_list, prompts)):
        print(f"处理视频 {i+1}/{len(video_list)}")

        extended = extender.extend_video(
            video=video,
            prompt=prompt,
            target_duration=target_duration
        )
        results.append(extended)

        # 避免触发速率限制
        time.sleep(5)

    return results

Comparing Veo 3.1 Video Extension with Other Solutions

There are several AI video generation options on the market. Here's how Veo 3.1's extend capability stacks up against other mainstream solutions:

Our Recommendations

• If you need the longest duration: Go with Veo 3.1; it supports up to 148 seconds.
• If you prioritize visual consistency: Choose Veo 3.1; it offers the best continuity.
• If you're on a budget: You can get better pricing through the APIYI (apiyi.com) platform.
• If you need quick results: The Veo 3.1 Fast version provides a much snappier response.

Veo 3.1 Video Extension FAQ

Veo 3.1 Video Extension API Error Handling

In real-world use, you might run into various errors. Here are some common ones and their solutions:

# 错误处理示例
def safe_extend_video(video, prompt, max_retries=3):
    """带重试机制的视频扩展"""
    for attempt in range(max_retries):
        try:
            operation = client.models.generate_videos(
                model="veo-3.1-generate-preview",
                prompt=prompt,
                video=video
            )

            while not operation.done:
                time.sleep(30)
                operation = client.operations.get(operation)

            return operation.result.generated_videos[0]

        except Exception as e:
            print(f"尝试 {attempt + 1} 失败: {e}")
            if attempt < max_retries - 1:
                time.sleep(60)  # 等待后重试
            else:
                raise

Veo 3.1 Video Extension API Use Cases

Short Video Creation

Extend an 8-second clip into a full short video of 60 seconds or more. This is perfect for:

• Douyin/TikTok content creation
• YouTube Shorts production
• Social media ad assets

Film & Video Previews

Quickly generate 1-2 minute concept videos for:

• Script visualization
• Storyboard animation
• Creative pitch presentations

Product Showcases

Create dynamic product showcase videos:

• 360° product rotations
• Usage scenario simulations
• Feature demo animations

Summary

Google Veo 3.1's video extension capability brings a breakthrough for AI video length:

1. 7-Second Incremental Mechanism: Each extension adds a fixed 7 seconds, sampling from the final 1 second of the previous segment to ensure seamless continuity.
2. 20-Iteration Limit: You can perform a maximum of 20 extension operations on a single project.
3. 148-Second Maximum Output: 8s original + 140s extension = 148s total duration.
4. Strict Input Requirements: It only supports MP4 videos generated by Veo at 24fps, with 720p or 1080p resolution.

If you're a developer or creator needing long-form video, mastering the Veo 3.1 extend API will seriously level up your production efficiency. We recommend checking out APIYI at apiyi.com to quickly test these effects and access reliable API services.

This article was written by the APIYI technical team. For more AI API tutorials, visit apiyi.com.