AI-Powered Fan Highlight Reels for stealth company powering fan engagement
Budget: $25 – $50 USD
I need an end-to-end AI workflow that can take the flood of fan-shot clips pouring in right after concerts and sporting events and turn them into share-worthy highlight reels without human screening.
Here’s the core of what the system must do:
• Ingest hundreds—even thousands—of vertical videos the moment they are uploaded.
• Rank every clip primarily on:
– Quality of the video
– Crowd excitement level
(Artist or player visibility is nice to have but not essential for the first release.)
• Autonomously pick the “Top 100 Moments” from the event and stitch them into a short, cohesive montage.
• Generate additional sponsor-branded cuts on demand (for example Heineken moments or VIP moments) by re-ordering or selectively replacing shots.
• Push the finished reel straight to TikTok via API within minutes of the final whistle or last encore.
You are free to combine computer-vision models, audio analysis, and heuristic scoring, as long as the final pipeline is reproducible and well documented. I’ll provide sample event footage and brand assets; you deliver:
1. Training and inference code (Python preferred)
2. A lightweight dashboard or CLI for running the pipeline on new events
3. Clear README plus model weights ready for deployment
The project is considered complete when I can hit one command after an event, receive a polished Top-100 montage plus any requested sponsor edits, and see the finished vertical video live on TikTok.
Here’s the core of what the system must do:
• Ingest hundreds—even thousands—of vertical videos the moment they are uploaded.
• Rank every clip primarily on:
– Quality of the video
– Crowd excitement level
(Artist or player visibility is nice to have but not essential for the first release.)
• Autonomously pick the “Top 100 Moments” from the event and stitch them into a short, cohesive montage.
• Generate additional sponsor-branded cuts on demand (for example Heineken moments or VIP moments) by re-ordering or selectively replacing shots.
• Push the finished reel straight to TikTok via API within minutes of the final whistle or last encore.
You are free to combine computer-vision models, audio analysis, and heuristic scoring, as long as the final pipeline is reproducible and well documented. I’ll provide sample event footage and brand assets; you deliver:
1. Training and inference code (Python preferred)
2. A lightweight dashboard or CLI for running the pipeline on new events
3. Clear README plus model weights ready for deployment
The project is considered complete when I can hit one command after an event, receive a polished Top-100 montage plus any requested sponsor edits, and see the finished vertical video live on TikTok.