🎬 Flow-Studio | Google Flow AI Studio

A production-ready Web Studio, CLI, and Automation Engine for Google Flow (flow.google.com).
Generate next-generation AI images and cinematic videos with real-time credit tracking, batch generation, and automated media downloading.

Features • Installation • Cookie Setup • CLI Usage • Web Studio • Supported Models • REST API

---

🌟 Key Features

• ⚡ Real-Time Credit Synchronization:

  • Live queries to Google Flow's /VideoFxService.GetCredits (nzlxg) and /AiSandbox.GetUserVideoCreditStatusAction (DviW9c).
  • Real-time balance badges, dynamic tier detection, and automatic credit deductions after each generation.

• 🎨 Multi-Model AI Image Engine:

  • Powered by Google's latest image architectures: Nano Banana 2.1 (Beluga), Nano Banana Pro (Gemini Pix 2), and Nano Banana 2 Lite (Harbor Seal / Imagen 3 variants).
  • Multiple aspect ratios: 16:9 (Landscape), 9:16 (Story/Reels), 1:1 (Square), 4:3, and 3:4.
  • Batch generation (1 to 4 images per prompt), negative prompts, and seed reproducibility.

• 🎥 Cinematic AI Video Engine:

  • Integration with Gemini Omni 1.1 Flash (Abra - ultra-fast video rendering with synchronized audio) and Veo 3.1 series (Fast, Lite, Quality).
  • 3 Flexible Generation Modes:
    1. Text-to-Video: Direct prompt-to-video cinematic rendering.
    2. Image-to-Video: Animate starting images (Start Frame).
    3. Start & End Frame Interpolation: Generate smooth transitions between two keyframe images.
  • Configurable video durations (4s, 6s, 8s, 10s), resolutions (720p, 360p), and aspect ratios (16:9, 9:16).
  • 1080p (1K Full HD) Upscaling: Retrieve and download high-bitrate 1080p upscaled videos directly from Google Flow.

• 💾 Automated Asset Pipeline:

  • Automatically downloads generated images and videos directly to local storage (downloads/).
  • Built-in Gallery browser with one-click folder exploration.

• 🖥️ Dual Interface (CLI & Web Studio):

  • Modern Web Studio: Dark-mode, glassmorphism UI built with FastAPI, TailwindCSS, and Alpine.js.
  • Powerful CLI Tool: Direct terminal workflows, automation scripting, and batch job processing.

• 🚀 Hybrid High-Performance Architecture:

  • Direct reverse-engineered Google Boq RPC / batchexecute client (flow_client.py) for lightweight, low-latency execution.
  • Optional headless Playwright automation worker (flow_worker.py) for browser canvas fallback.

---

📋 Supported Models & Specifications

1. Image Generation Models

Model Name	Internal Key	Badge	Description
Nano Banana 2.1	BELUGA	🍌 Latest	Balanced next-generation model with high fidelity and prompt adherence.
Nano Banana Pro	GEM_PIX_2	🍌 Pro	Flagship high-detail model supporting upscaling and ultra-sharp textures.
Nano Banana 2 Lite	HARBOR_SEAL	🍌 Lite	Ultra-fast lightweight model, minimal credit consumption.

Supported Aspect Ratios: 16:9, 9:16, 1:1, 4:3, 3:4
Batch Output: 1 to 4 images per generation

---

2. Video Generation Models

Model Name	Internal Key	Badge	Audio Support	Description
Omni 1.1 Flash	abra	🔊 Omni Flash	Yes	High-speed Gemini Omni model generating fluid motion and audio.
Veo 3.1 - Fast	veo_3_1_fast	🔊 Fast	Yes	Quick rendering speed, ideal for rapid iterations and testing.
Veo 3.1 - Lite	veo_3_1_lite	🔊 Lite	Yes	Resource-efficient variant with reduced credit consumption.
Veo 3.1 - Quality	veo_3_1_quality	🔊 Cinematic	Yes	Google DeepMind's flagship cinematic video generation engine.

Video Specifications:

• Aspect Ratios: 16:9 (Horizontal) or 9:16 (Vertical Reels/TikTok)
• Durations: 4s, 6s, 8s (Default), 10s
• Resolutions: 720p (Default), 360p
• Generation Modes: Text-to-Video, Image-to-Video (Start Frame), Transition (Start & End Frame)

---

🚀 Quick Start

1. Clone Repository & Setup Environment

# Clone the repository
git clone https://github.com/NePreby/Flow-Studio.git
cd Flow-Studio

# Create a virtual environment
python -m venv venv

# Activate virtual environment
# Windows (cmd/PowerShell):
venv\Scripts\activate
# Linux/macOS:
source venv/bin/activate

# Install required dependencies
pip install -r requirements.txt

(Optional: If utilizing Playwright browser worker fallback)

playwright install chromium

---

🔑 Google Flow Cookie Setup

To access your Google Flow credits and authentic generation pipeline, configure your session cookie:

1. Open your browser and navigate to https://flow.google.com.
2. Log into your Google account.
3. Press F12 (or Right-click -> Inspect) to open Developer Tools.
4. Go to the Network tab and refresh the page (F5).
5. Select the first network request to flow.google.com.
6. In Request Headers, locate the Cookie: entry and copy its entire string.
7. Save the cookie using either the Web UI or CLI:

Via CLI:

python cli.py set-cookie "YOUR_COPIED_COOKIE_STRING"

Via Web UI: Navigate to the Cookie Settings tab in the Web Dashboard, paste the cookie, and click Save & Test Connection.

---

💻 CLI Usage Guide

Run commands directly from your terminal using cli.py:

Check Account Balance & Credits

python cli.py credits

List Available Models & Formats

python cli.py models

Generate Images

# Basic image generation (2 images, 16:9 ratio)
python cli.py image "Cyberpunk cityscape with neon reflections on wet asphalt, 8k cinematic"

# Advanced customization
python cli.py image "A majestic mechanical tiger in an overgrown botanical greenhouse" \
  --model "Nano Banana Pro" \
  --ratio 16:9 \
  --count 4 \
  --seed 42 \
  --negative "blurry, low quality, artifacts" \
  --output ./my_renders \
  --download

Generate Videos

# Text-to-Video with automatic 1080p (1K) upscaled download
python cli.py video "FPV drone cinematic flythrough of a glowing crystal cave" \
  --model "Omni 1.1 Flash" \
  --ratio 16:9 \
  --resolution 720p \
  --quality 1080p \
  --duration 8 \
  --count 1 \
  --download

# Image-to-Video (Start Frame)
python cli.py video "The camera orbits around the character as wind blows gently" \
  --start-image "https://example.com/start_frame.jpg" \
  --duration 6

# Start and End Frame Interpolation
python cli.py video "Smooth morph transition from sunrise to twilight" \
  --start-image "https://example.com/day.jpg" \
  --end-image "https://example.com/night.jpg" \
  --duration 8

# Upscale an existing video to 1080p Full HD by media ID
python cli.py upscale "c5d19a3b-7821-4f11-9e2a-18b9f39031d2"

---

🌐 Web Studio UI

Launch the Web Interface

Windows 1-Click Launch: Double-click run_web.bat, or execute:

python -m uvicorn app:app --host 127.0.0.1 --port 7860 --reload

The Web Studio will be accessible at: http://localhost:7860

Highlights of the Web Studio:

• Interactive Prompt Canvas: Select models, aspect ratios, seeds, and image counts with visual badges.
• Visual Drag-and-Drop Uploader: Effortlessly upload Start Frame and End Frame images for video generation.
• Live Output Player & Gallery: Preview videos with audio, inspect high-res images, and trigger immediate downloads.
• Native File Explorer Integration: Open local download folders directly via native system calls.

---

🔌 API Endpoints

The FastAPI server (app.py) exposes full RESTful endpoints for integration into third-party applications:

Method	Endpoint	Description
GET	/api/credits	Fetch live Google Flow balance and account tier.
GET	/api/models	List all available image & video models with configurations.
GET	/api/config	Retrieve current application configuration and masked cookie.
POST	/api/cookie	Update and test Google Flow session cookie.
POST	/api/generate/image	Submit an image generation request.
POST	/api/generate/video	Submit a video generation request (Text/Image/Keyframe).
POST	/api/upscale-video	Upscale and download video in 1080p (1K Full HD) quality.
POST	/api/upload	Upload local images for Start/End frame video generation.
POST	/api/download	Download remote media assets to the local server storage.
GET	/api/gallery	Retrieve metadata of locally stored assets in downloads/.
POST	/api/open-downloads	Launch the local file explorer at the download directory.

---

📁 Project Architecture

Flow-Studio/
├── app.py                # FastAPI application & RESTful routing
├── cli.py                # Command-line interface with argparse & Rich formatting
├── flow_client.py        # Core reverse-engineered Google Flow RPC / Boq client
├── flow_worker.py        # Playwright headless browser fallback worker
├── config.py             # Settings, credential storage, and directory management
├── config.json           # Local persistent configuration file
├── requirements.txt      # Python dependencies
├── run_web.bat           # 1-Click launcher script for Windows
├── web/
│   └── index.html        # Modern dark-mode Web Studio interface
├── downloads/            # Local directory for generated and downloaded media
└── uploads/              # Local storage for uploaded keyframe references

---

⚙️ Configuration File (config.json)

The application automatically persists configuration in config.json:

{
  "cookie": "YOUR_FLOW_COOKIE",
  "downloads_dir": "downloads",
  "auto_download": true,
  "default_image_model": "Nano Banana 2.1",
  "default_video_model": "Omni 1.1 Flash",
  "default_duration": 8,
  "default_count": 2
}

---

⚠️ Disclaimer & Notes

• Disclaimer: This tool is an independent open-source project intended for research, personal automation, and educational purposes. It is not affiliated with, endorsed by, or sponsored by Google.
• Account Safety: Keep your session cookies private and never commit config.json containing sensitive cookies to public repositories.

---

📄 License

Distributed under the MIT License. See LICENSE for more details.