This project provides a standalone system for generating AI-driven video content via the Google Flow API, designed to automate the creation of professional-quality videos from concept to final output. It targets engineers, researchers, and power users seeking to streamline video production, offering end-to-end capabilities including narration, branding, and YouTube-ready formatting, thereby eliminating manual effort.

How It Works

The system employs a Python agent (FastAPI) that communicates with a Chrome extension (MV3 Service Worker) via WebSockets. This extension acts as a crucial bridge, handling authentication, reCAPTCHA solving, and proxying API calls to Google Flow. A core innovation is its "reference image system," which ensures visual consistency across scenes by generating and reusing distinct reference images for characters, locations, and props. Scene prompts then focus solely on actions, leveraging these references for appearance. The pipeline orchestrates story conception, entity definition, reference image generation, scene composition, video clip creation, text-to-speech narration, concatenation, thumbnail generation, and YouTube metadata creation.

Quick Start & Requirements

• Primary install: Execute ./setup.sh for a one-command setup that checks and installs Python 3.10+, pip, ffmpeg, ffprobe, and Chrome, then creates a virtual environment and installs dependencies.
• Manual setup: Ensure Python 3.10+, ffmpeg, and Chrome are installed, then run pip install -r requirements.txt.
• Prerequisites: Python 3.10+, ffmpeg, ffprobe, Google Chrome browser.
• Environment: All bash scripts assume a Unix shell; Windows users should utilize WSL (wsl --install) or Git Bash.
• Relevant Links: Google Flow UI: https://labs.google/fx/tools/flow

Highlighted Details

• End-to-End Video Production: Generates complete YouTube-ready videos, including visual consistency, voice-cloned narration, and SEO-optimized metadata.
• Visual Consistency: The reference image system maintains identical character and object appearances across diverse scenes and lighting conditions.
• Chrome Extension Dashboard: Offers real-time monitoring of video generation progress, API request logs, and token status.
• AI Agent Skills: Provides a set of CLI-agnostic skills for automating the entire workflow, from project creation to YouTube uploads.

Maintenance & Community

• Community: A shared community for FlowKit and Flowboard is available on Facebook: facebook.com/groups/flowkit.flowboard.community.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license permits commercial use. However, functionality is contingent on the Google Flow API and the Chrome extension, which may have separate terms of service.

Limitations & Caveats

The system is tightly coupled to the Google Flow API and requires a continuously running Chrome browser with the provided extension. Setup involves specific software versions (Python 3.10+, ffmpeg) and a Unix-like shell environment. Functionality like 4K upscaling is designated as "TIER_TWO only," implying potential cost or access limitations. The reliance on reCAPTCHA solving via the extension can introduce points of failure.