Akash Nath

This project aims to detect and highlight deceptive design patterns (dark patterns) on websites. Dark patterns are UI/UX strategies used to manipulate users into making unintended decisions, such as forced subscriptions, hidden costs, misleading urgency, or FOMO tactics. The Problem: Many websites use psychological tricks to increase conversions at the cost of user trust and autonomy. Users often don’t realize they are being manipulated until it’s too late. The Solution: This project automatically scans web pages, detects potential dark patterns using a deep learning model, and visually highlights them. The AI classifies deceptive elements with confidence scores, using color-coded warnings (green for low certainty, red for high certainty) to help users identify manipulative tactics at a glance. Impact: > Empowers users by making manipulative design tactics transparent. > Increases awareness of deceptive UI/UX practices. > Encourages ethical web design by exposing dark patterns. In short, this tool helps fight against deceptive digital practices and promotes a more transparent, user-friendly internet.

🚀 The Problem It Solves Video editing is traditionally a time-consuming, highly technical, and often tedious process. Whether you're a content creator trying to keep up with the relentless pace of social media or a professional editor bogged down by the initial rough cut, hours are lost simply sifting through raw footage, finding the right moments, syncing clips to audio, and applying basic transitions or color grading. VibeCut completely reimagines this workflow by turning video editing into a conversational, intent-driven experience. By abstracting away the complex mechanics of traditional NLEs (Non-Linear Editors) behind a team of specialized AI agents, VibeCut gives users their time back and lets them focus purely on creative storytelling. What it can be used for: Rapid Social Media Content Creation: Creators can upload a folder of raw b-roll, provide a simple prompt (e.g., "Create a high-energy travel reel using this script..."), and VibeCut will automatically select the best clips, generate text overlays, and match the cuts to the music. Automated Rough Cuts: Video editors and filmmakers can use it to instantly generate a baseline timeline from hours of raw footage, saving them from the grueling, repetitive task of initial clip sorting and assembly. Accessible Storytelling: Individuals or small businesses with zero video editing experience or technical knowledge can produce high-quality, engaging promotional videos just by describing the "vibe" they want to achieve. How it makes existing tasks easier and incredibly fast: Intelligent Media Analysis: Instead of manually scrubbing through hours of clips, VibeCut's multi-agent system uses Vision and Speech AI to automatically analyze, transcribe, and tag your footage based on sentiment, action, and visual content. Context-Aware Assembly: The Edit Planner and Preset Intelligence agents translate your natural language prompt into precise timeline operations. They automatically apply the correct color profiles, motion effects, and typography that fit the exact mood of your request. Real-Time, Conversational Iteration: Making changes is as simple as chatting with the AI. You can simply ask the orchestrator to "make the pacing punchier" or "swap the second clip with something more cinematic," and the React-based timeline updates dynamically in real-time. Multi-Format Generation: With built-in aspect ratio intelligence, repurposing content for different platforms (e.g., converting a 16:9 YouTube video into a 9:16 TikTok/Reel) is handled natively by the agentic pipeline without the need for manual reframing. 🏗️ Architecture VibeCut is built on a Python/FastAPI backend and a React frontend connected over WebSocket for real-time updates. The core of the system is a 10-agent agentic pipeline powered by Google Gemini, where each agent is a specialist responsible for one stage of the editing process — from media ingestion and scene analysis, all the way to music selection and final FFmpeg export. The user's natural language prompt enters the Orchestrator, which coordinates the full pipeline. Media is first analyzed by Ingestion, Speech, and Vision agents to extract metadata, transcripts, and visual context. This indexed context flows into the Clip Retrieval agent for semantic search. The Edit Planner then converts the user's intent into precise timeline operations, collaborating with the Preset Intelligence, Music, and Gen Media agents before handing off to the Editing Execution agent for the final render. image 🌟 Example Workflow Open the editor. Upload video clips or set folder_path when creating a project. Type: > Create a high-energy travel reel using this script: "Welcome to paradise. The beaches are stunning. The food is incredible. Let's explore together." Watch the timeline assemble live with clips, text overlays, color presets, and music.