crisng95/flowkit

What it solves

Flow Kit is a standalone system designed to automate the end-to-end production of AI-generated videos using the Google Flow API. It solves the challenge of maintaining visual consistency across multiple scenes—ensuring characters, locations, and objects look the same throughout a video—while automating the tedious process of authentication, reCAPTCHA solving, and API proxying.

How it works

The system uses a Python-based agent (FastAPI) that communicates via WebSockets with a Chrome extension. This extension acts as a bridge to the Google Flow interface, handling authentication and proxying API calls.

The core of its consistency is a Reference Image System: users define "entities" (characters, locations, or assets) with appearance descriptions. The system generates a single reference image for each entity, which is then passed as a visual input to every scene generation, allowing the AI to maintain the same face, clothing, and features across different actions and settings.

Who it’s for

It is built for content creators and AI agents who want to produce high-quality, consistent narrative videos (such as YouTube stories) without manually managing every frame or prompt.

Highlights

• End-to-End Pipeline: Automates the entire flow from story concept and entity definition to reference images, scene images, 8-second video clips, TTS narration, and final concatenation.
• Visual Consistency: Uses a dedicated reference system to keep characters and environments stable across dozens of scenes.
• Automated Browser Bridge: A Chrome extension handles the complex parts of interacting with Google Flow, including token capture and reCAPTCHA solving.
• Ops Console: Includes a local React dashboard for monitoring pipeline progress, reviewing AI-generated scenes, and managing projects.
• AI Agent Integration: Provides "skills" (slash commands) that allow other AI agents to drive the entire video production process.