AI Video Clipper
A locally running AI-powered video clipping application that analyzes long videos, identifies engaging highlights, and generates vertical reels ready for preview and download.
- Year
- 2026
- Role
- UI/UX Designer & Full Stack Developer
- Client
- Personal Project
- Category
- AI / Full Stack Development
AI Video Clipper is a full-stack application designed to simplify the process of turning long-form videos into short-form content. Users can upload a local video or provide a public YouTube URL, choose their preferred reel style, and generate clips using a locally running AI processing pipeline. The application combines speech transcription, AI-assisted highlight selection, and FFmpeg-based video rendering to produce vertical videos. Users can preview and download generated clips, track processing progress, and revisit previous projects through a persistent project history. The system is designed to run locally, keeping video processing and language-model inference on the user's machine rather than relying on paid cloud AI APIs.
Creating short-form videos from long recordings requires manually reviewing footage, identifying engaging moments, trimming clips, and formatting videos for vertical platforms. This process can be repetitive and time-consuming.
Build a locally running application that automates highlight discovery and reel generation while giving users control over the final clips and video presentation style.
Designed the user experience and developed the frontend, backend API integration, local AI processing workflow, and video-generation pipeline. Implemented video uploads, YouTube URL ingestion, processing progress, project history, clip previews, and downloads.
1. Designed a dark, modern interface for video creation and project management. 2. Built the React and TypeScript frontend. 3. Developed the ASP.NET Core backend to manage processing jobs. 4. Integrated Python-based transcription and local LLM-assisted highlight selection. 5. Connected FFmpeg for vertical video rendering and yt-dlp for YouTube downloads. 6. Tested processing, playback, downloads, and local service lifecycle.
Coordinating multiple local services and technologies while maintaining a reliable end-to-end workflow. Additional challenges included handling video downloads, preserving audio during rendering, managing long-running processing jobs, and making generated clips accessible through the frontend.
Created a shared processing workflow in which both uploaded videos and downloaded YouTube videos are passed to the same Python engine. Used ASP.NET Core to orchestrate jobs, Ollama for local language-model inference, faster-whisper for transcription, and FFmpeg for rendering and media processing.
Built an integrated local video-clipping application with a React interface, .NET backend, Python AI engine, and FFmpeg rendering pipeline. The application supports video ingestion, AI-assisted clip generation, and browser-based clip playback and downloads.
- Local video upload and public YouTube URL input.
- AI-assisted highlight detection.
- Speech transcription using faster-whisper.
- Local LLM inference through Ollama.
- Multiple vertical reel styles: blurred background, full frame, and focused reel.
- FFmpeg-powered video rendering.
- Video preview, playback, and downloads.
- Project history and processing status tracking.
- Local-first processing without cloud AI dependencies.