Leveraging SOTA MultiModal AI for Video Understanding- An Iterative Approach to Replicating Viral Success on TikTok
???? Data Collection:
- Apify API Integration: Scrapes #TikTok data based on your query
- Recursive Searches: Finds related searches and aggregates them
- Automatic downloading of the videos found for each search.
???? Automated Transcription:
- Audio & Vocal Separation: ffmpeg & Spleeter make it happen ????.
- #Openai's #Whisper for Transcription: Accurate and efficient.
????Video Understanding with #Vicuna (The Special Part!):
- Vicuna 7b video understanding Model "watches" the videos ????.
- Paired with Transcriptions: A unique combo of insights ????.
- Custom Insights: Tailor the analysis to your goals.
???? Automated Report Generation:
- Aggregate analysis: Using the transcripts and video understanding
- Markdown Reports: Neat and strategic final report based on all learnings.
????️ Requirements & Usage:
- Apify API Key: For data collection.
- OpenAI API Key: For Vicuna & GPT-3.
- Google Colab+: To run it on an A100 ????.
???? Key Benefits:
- ???? Fully automated analysis
- ???? Unlocks strategic marketing insights using AI
- ???? Automated discovery of what makes themes of videos viral
- ???? Identifies key factors for TikTok success
-???? Evolving understanding of viral patterns
- ???? Structured report format for decision makers
- ???? Custom analysis based on goals
** You must run the script in order and follow the directions exactly. Because of its complexity, you may run into errors. I'll try to support as possible, but this script is experimental.```
Here is the colab notebook if you want to try:
<https://colab.research.google.com/drive/1vS9lJvgytPVvIj_PsMbs2Tc5BgiVCbp-#scrollTo=FA4nUHX-cpPB>
