Skip to content

UnknownX

Ctrl + F for Videos and Creative Direction - Talk with agent about visuals on your mind and fetch them to convert int your product/character frames using NBlite and animation using Omni Flash.

Code Video Cerebral Valley

Core trick

A local Gemma model turns spoken requests into searches over local video footage, then passes matched frames to an image model for recasting.

Tags

visioncreative/music

Description

Ctrl + F for Videos and Creative Direction - Talk with agent about visuals on your mind and fetch them to convert int your product/character frames using NBlite and animation using Omni Flash.
dinadhilip/Vision_Room_Alt_Gemma★ 00 forksPythonCreated Jul 11, 2026 · last push Jul 11, 2026

README excerpt

Update : New Loom Video is out (Code is in a branch)!

# Vision Room Alt Gemma

Vision Room is an edge-to-cloud conversational video studio. It uses a local Gemma model to orchestrate searches over your local video footage, Nano Banana 2 Lite (NB2 Lite) to cast subjects into matching frames, and Omni Flash to synthesize a final video.

## Prerequisites & Environment Variables

Before running the application, make sure to set the required API keys and endpoint URLs in your `.env` file at the root of the project:

```env
VISION_ROOM_LITERT_BASE_URL="http://127.0.0.1:8000"
VISION_ROOM_LITERT_MODEL="gemma-4-local"
VISION_ROOM_NB2_LITE_API_KEY="your_actual_gemini_api_key_here"
# Add your Omni Flash API key when ready for video synthesis
# VISION_ROOM_OMNI_FLASH_API_KEY="..."
```

*Note: If an API key is missing or a cloud provider call fails, the bridge API will automatically fall back to local deterministic demo stubs so your workflow isn't blocked.*

## How to Run the Application

The application requires two servers running simultaneously. Open two separate terminal windows inside the project root (`/Users/vyshnav/Documents/Vision room`).

### 1. Start the Local Model Server (Terminal 1)
Run the local `litert-lm` server to provide the Gemma-4-local OpenAI-compatible endpoint:
```bash
./.venv/bin/litert-lm serve --port 8000
```

### 2. Start the Bridge API & UI (Terminal 2)
Run the main FastAPI bridge which hosts the UI, handles chat routing, and orchestrates the backend skills:

Team