vercel/ai · error · AISDKError

VERTEX_VIDEO_GENERATION_ERROR

VERTEX_VIDEO_GENERATION_ERROR

Error message

No videos in response. Response: ${JSON.stringify(finalOperation)}

What it means

buildCompletedResult in the Vertex video model throws an AISDKError named VERTEX_VIDEO_GENERATION_ERROR when the finished long-running operation's response contains no `videos` array (missing or empty). It wraps the entire serialized operation in the message so the raw API payload can be inspected.

Source

Thrown at packages/google-vertex/src/google-vertex-video-model.ts:280

    warnings: SharedV4Warning[];
    currentDate: Date;
  }): {
    status: 'completed';
    videos: Array<
      | { type: 'base64'; data: string; mediaType: string }
      | { type: 'url'; url: string; mediaType: string }
    >;
    warnings: SharedV4Warning[];
    providerMetadata: SharedV4ProviderMetadata;
    response: {
      timestamp: Date;
      modelId: string;
      headers: Record<string, string> | undefined;
    };
  } {
    const response = finalOperation.response;
    if (!response?.videos || response.videos.length === 0) {
      throw new AISDKError({
        name: 'VERTEX_VIDEO_GENERATION_ERROR',
        message: `No videos in response. Response: ${JSON.stringify(finalOperation)}`,
      });
    }

    const videos: Array<
      | { type: 'base64'; data: string; mediaType: string }
      | { type: 'url'; url: string; mediaType: string }
    > = [];
    const videoMetadata: Array<{
      gcsUri?: string | null | undefined;
      mimeType?: string | null | undefined;
    }> = [];

    for (const video of response.videos) {
      if (video.bytesBase64Encoded) {
        videos.push({
          type: 'base64',

View on GitHub (pinned to 69428b1f8b)

Solutions

Inspect the serialized operation in the error message for `raiMediaFilteredReasons` or similar filtering fields and adjust the prompt.
Retry generation with a modified prompt that avoids safety triggers.
Upgrade @ai-sdk/google-vertex in case the response schema handling is outdated.
Check the model/region supports video output before retrying.

Defensive patterns

Strategy: try-catch

Try / catch

import { AISDKError } from 'ai';
try {
  const result = await generateVideo({ model: vertex.video('veo-3.0-generate-001'), prompt });
} catch (e) {
  if (AISDKError.isInstance(e) && e.name === 'VERTEX_VIDEO_GENERATION_ERROR' && e.message.includes('No videos in response')) {
    // inspect operation payload for safety-filter reasons, adjust prompt, retry
  } else throw e;
}

Prevention

Check the operation payload for raiMediaFilteredReasons and pre-screen prompts against Veo safety policies.
Pin and regularly update @ai-sdk/google-vertex to track Veo response-schema changes.
Verify the chosen region supports video output.

When it happens

Trigger: Polling doStatus for a completed Veo generateVideos operation whose finalOperation.response lacks videos — e.g. the API returned success but filtered out all output (safety/content filtering), or an unexpected response shape after an API change.

Common situations: Veo silently dropping generated video due to prompt safety filters; malformed/empty operation.response from the Vertex API; outdated provider version vs. changed Veo response schema; region where video output is unavailable.

Related errors

AI-assisted analysis of vercel/ai@69428b1f8b (2026-08-30). Data as JSON: /api/errors/b3159ba5e0e9e384. Report an issue: GitHub.