[Feature Request] Add prompt optimization profile for Google Gemini Omni video generation models

[arguswaikhom]

Description

I propose adding a dedicated tool profile and routing rule for the Google Gemini Omni model family (e.g., gemini-omni-flash).

Gemini Omni introduces a native, any-input-to-video architecture that processes text, images, audio, and reference videos within a single conversational pipeline. Having prompt-master generate structured prompts optimized for this unified multi-modal input and multi-turn conversational video editing would be incredibly valuable for precise creative control.

Additionally, because Gemini Omni generates a maximum of 10 seconds per individual video clip, the profile should support multi-clip sequencing. If a user requests a prompt for a longer duration (e.g., 30 seconds), prompt-master should intelligently break down the request and generate a series of sequential 10-second clip prompts that can be stitched together seamlessly while maintaining narrative and visual continuity.

Use Case

A dedicated Gemini Omni profile will allow users to:

• Handle Native 10-Second Constraints: Seamlessly chunk a long-duration scene prompt into sequential, 10-second structural prompt blocks (e.g., Clip 1, Clip 2, Clip 3) optimized for multi-turn generation.
• Leverage Mixed-Modality Context: Structure prompts that explicitly direct the model on how to weigh combined inputs (e.g., a voiceover track, a reference image, and text style descriptions).
• Optimize for Multi-Turn Editing: Craft instruction layers tailored for sequential video manipulation (e.g., changing lighting, swapping objects, or altering angles in subsequent conversational turns while maintaining continuity).
• Ground in Physics & Motion: Utilize specialized keyword framing that taps into Omni's native understanding of physical dynamics and kinetic motion to reduce simulation anomalies.