Veo 3.1 Fast video extension returns HTTP 400 using original generated video URI

Hello Google team,

I can successfully generate and download an 8-second, 720p,
9:16 video using veo-3.1-fast-generate-preview.

However, extending that video fails during models.generateVideos
with HTTP 400. No extension operation ID was saved.

Environment:

  • Gemini API through Google AI Studio
  • Node.js v24.13.1
  • @google/genai v2.28.0

The original generation operation is:
models/veo-3.1-fast-generate-preview/operations/0htw6nr8nbgd

A status lookup confirms that the original operation completed
successfully and returned one video with a URI.

The extension uses that original URI:

await client.models.generateVideos({
model: “veo-3.1-fast-generate-preview”,
source: {
prompt: “Continue the scene. The creator says: Let’s take a closer look together.”,
video: {
uri: originalVideoUri
}
},
config: {
numberOfVideos: 1,
resolution: “720p”
}
});

The prompt above is shortened to show the request structure.

The parent video was generated less than one hour before the
extension attempt.

Our adapter unfortunately discarded the explanatory error message.
We have since improved diagnostics and stopped further submissions.

Questions:

  1. Is extension of 9:16 videos supported by this model?
  2. Is source.video.uri the correct input with SDK v2.28.0?
  3. Are additional configuration fields or project permissions required?
  4. Can you investigate the rejection or advise how to recover its
    error details?

Thank you.

I am also trying to extend a video.

The documentation here suggests that both 16:9 and 9:16 are supported, but I get this response from the API (POST to https://generativelanguage.googleapis.com/v1beta/models/veo-3.1-generate-preview:predictLongRunning):

{“error”:{“code”:400,“message”:“Aspect ratio of the input video must be 16:9, but got: 9:16”,“status”:“INVALID_ARGUMENT”}}

I am using

  • SDK: @google/genai, models.generateVideos

I notice that you must also specify the aspectRatio in the request to the extension API call

So this worked as long as I included the aspectRatio:
parameters: {
sampleCount: 1,
resolution: “720p”, “aspectRatio”: “9:16”
}

The doc implies you don’t need it, so I guess just the documentation needs updating:

Hello,

Thank you for reaching out, and thank you @Daniel_Glenn for sharing the working configuration.

@Grace_Ngugi — here are the answers to your questions:

  1. 9:16 video extension support: Yes, extending 9:16 (vertical) 720p Veo-generated videos is supported.

  2. Why HTTP 400 occurred (config.aspectRatio requirement): When aspectRatio is omitted from config in models.generateVideos, the request defaults to "16:9" and validates that the input video matches that aspect ratio, returning 400 INVALID_ARGUMENT: "Aspect ratio of the input video must be 16:9, but got: 9:16". Explicitly including aspectRatio: "9:16" in config alongside resolution: "720p" resolves the error:

    const operation = await client.models.generateVideos({
      model: "veo-3.1-fast-generate-preview",
      source: {
        prompt: "Continue the scene. The creator says: Let's take a closer look together.",
        video: {
          uri: originalVideoUri,
        },
      },
      config: {
        numberOfVideos: 1,
        resolution: "720p",
        aspectRatio: "9:16",
      },
    });
    
    
  3. Input format and permissions: Passing the prior operation’s generated video object (operation.response.generatedVideos[0].video or { uri: originalVideoUri } from a Veo generation completed within the last 2 days) is supported in @google/genai, and no additional project permissions are needed if initial Veo generation succeeds.

  4. Recovering error details: In @google/genai, wrap the client.models.generateVideos(...) call in a try / catch block and log error.status and error.message (which contains the JSON error payload with error.code, error.message, and error.status).

Please also note that per the Gemini API deprecations page and Veo documentation, veo-3.1-generate-preview and veo-3.1-fast-generate-preview were shut down on October 22, 2026, with gemini-omni-1.1-flash listed as the recommended replacement in the Gemini API. For ongoing video generation and video extension workflows, we recommend migrating to Gemini Omni Flash (gemini-omni-1.1-flash) via the Interactions API, or to the generally available Veo 3.1 models on the Gemini Enterprise Agent Platform.