Execute automatic activation for all google vertex ai multimodal operations operations.
ReadWriteEditGrepGlobBash(general:*)Bash(util:*)
Vertex AI Media Master
Overview
Multimodal media operations on Google Cloud Vertex AI covering video understanding, audio generation, image creation, and marketing campaign automation. This skill orchestrates Gemini 2.5 Pro/Flash, Imagen 4, and Lyria models to process, analyze, and generate rich media assets.
Prerequisites
- Google Cloud project with Vertex AI API enabled
google-cloud-aiplatform Python SDK installed (pip install google-cloud-aiplatform[vision,audio])GOOGLE_CLOUD_PROJECT and GOOGLE_APPLICATION_CREDENTIALS environment variables set- Service account with
roles/aiplatform.user permission
- Sufficient quota for target models (Gemini 2.5 Pro: 2M tokens/min; Imagen 4: 100 images/min)
Instructions
- Initialize the Vertex AI client with the target project and region (
us-central1 recommended for model availability).
- Select the appropriate model for the task:
- Video analysis: Gemini 2.5 Pro (up to 6 hours at low resolution, 2 hours at default).
- Image generation: Imagen 4 for highest quality stills; Gemini 2.5 Flash Image for interleaved text+image output.
- Audio generation: Lyria for music composition and background tracks.
- Campaign automation: Gemini 2.5 Pro for multi-asset generation from a single prompt.
- Prepare input media: upload source files to Cloud Storage (
gs:// URIs) or provide local paths for smaller assets.
- Construct the generation request with explicit parameters (aspect ratio, duration, number of outputs, style constraints).
- Execute the request and capture response objects containing generated media bytes or analysis text.
- Post-process outputs: save generated images/audio to the target directory, extract structured insights from video analysis, or compile campaign asset bundles.
- Validate results against brand guidelines or schema expectations before delivery.
Output
- Generated image files (PNG/JPEG) from Imagen 4 or Gemini Flash Image
- Audio files (WAV/MP3) from Lyria model for background music, voiceovers, or sound effects
- Video analysis reports: scene breakdowns, key-moment timestamps, transcript text, marketing-insight summaries
- Campaign asset packages: hero images, social media graphics, ad copy, email marketing text, and video scripts
- Structured JSON metadata for each generated asset (model used, prompt, parameters, cost estimate)
Error Handling
| Error |
Cause |
Solution |
PermissionDenied on Vertex AI API |
Service account lacks aiplatform.user role |
Grant the require
Ready to use 003-jeremy-vertex-ai-media-master?
|