← Back
VIBE AI — YouTube Idea Generator
Transforming comments and trends into trend-informed video ideas.
Overview
VIBE AI ingests YouTube comments, metadata, and trends to produce tailored video concepts, titles, and prompts. The system combines classification models (RNN, BERT) to mine suggestions, and a fine-tuned GPT-2 to generate idea variations aligned with creator style and current trends.
Methodology & Models
- Suggestion mining: RNN + BERT to filter and classify high-quality suggestion comments.
- Generation: GPT-2 fine-tuned on YouTube metadata, comments, and trend data.
- Data sources: YouTube Data API, Selenium scraping, Kaggle datasets, Google Trends.
Tools & Role
Role: data collection, model fine-tuning, evaluation, and prototype UI. Tools: PyTorch, Hugging Face Transformers, Selenium, GPT-2, BERT, RNNs, and standard ML tooling.
Outcome
VIBE AI provides creators with trend-aligned, high-quality idea suggestions, improving ideation throughput and helping smaller creators generate timely, audience-relevant content.