Channel-IQ
AI video-processing platform that turns long-form YouTube videos into short clips and publishing assets.
STATUS: Final-year project
01. Overview
AI video-processing platform that turns long-form YouTube videos into short clips and publishing assets.
02. Interface / Screenshots
03. Architecture
Interactive Architecture Walkthrough — a visualization of the implemented architecture/workflow.
Interactive Architecture Walkthrough
A pipeline that downloads video, transcribes audio, selects highlights, renders clips, and prepares upload metadata. [Visualization of the implemented workflow.]
04. Problem
Long-form videos take significant manual effort to review, trim, caption, and prepare for short-form publishing.
05. Solution
Built a video-processing workflow that downloads source footage, transcribes audio, identifies highlights, applies face-aware framing, renders vertical clips, and prepares SEO metadata for upload and publishing.
AI focus
- -OpenAI highlight selection from timestamped transcript segments
- -YOLO face detection and OpenCV crop logic
- -Face-aware framing for active-speaker and multi-face clips
06. Key Features
- >YouTube video input and download
- >FFmpeg audio extraction
- >Whisper transcription with timestamp handling
- >OpenAI highlight selection
- >Three generated short-form clips
- >YOLO face detection and face-aware framing
- >16:9 to 9:16 conversion
- >Caption, title, description, and tag generation
- >OAuth authentication and YouTube upload
- >Scheduled publishing
07. Engineering Challenges
Combined Whisper output with timestamp-aware processing before clip rendering.
Applied face-aware framing before FFmpeg rendering.
08. Outcome
Automated the pipeline from long-form video ingestion to generated vertical clips and publishing assets; the GPU workflow reduced a roughly one-hour video from 3-4 hours of manual editing to under 25 minutes.