Keep pulling the thread on Katie Nguyen.
Google Cloud's MCP servers are used to access the Gemini text-to-speech model, the Nano Banana image generation model, and the Lyria music generation model.
Google's Vio model performs image-to-video generation and can use a starting image as a baseline to maintain character consistency in subsequent video scenes.
Using Google's agentic framework, developers can create a separate 'image evaluator agent' that uses an LLM as a judge to compare generated media against the original prompt for adherence.
The multimodal capabilities of Google's Gemini model allow it to analyze images and fact-check them against a series of questions to ensure alignment with a prompt.
Google Cloud provides an Agent Development Kit (ADK) framework for building agentic workflows.
In an agentic workflow, Google's Gemini model can maintain memory and track context across multiple tasks to ensure consistency.
The demonstrated agentic framework uses Google's Nano Banana 2 model for character image generation from text descriptions.
Google's Nano Banana 2 model is capable of preserving details from natural language prompts when generating image outputs.
The demonstrated agentic workflow uses Google's Lyria model to generate and layer audio into videos.
Google has launched a new text-to-speech model, version 3.1, which is part of the Gemini family.
The demonstrated agent utilizes the Google Gen AI SDK to programmatically generate videos.
Google Cloud provides MCP servers built to access its Gen Media models.