Emma Huang and Louis Huang: Multimodal AI for Detecting Deepfake Videos
Multimodal AI for Detecting Deepfake Videos.
Emma Huang and Louis Huang developed a deep-learning model that analyses both audio and video to improve deepfake detection. Synthetic media can leave inconsistencies across modalities, so comparing what is seen with what is heard can strengthen detection.
Innovation mindset: When deception is designed to fool one sense, cross-check independent channels for signals that do not agree.
Official project title: A Novel Audio-Video Multimodal Deep Learning Model for Improved Deepfake Detection To Combat Disinformation (Behavioral and Social Sciences).
IP note
No public IP record; any filing made around the May 2025 fair would not yet be published under the 18-month rule — register check due from late 2026.
Recognition
Regeneron ISEF 2025 — Second Award ($2,400).
What this means for a student inventor in India
[Editor: 120–200 words connecting this story to the Indian route — Section 6 (any true inventor may apply), provisional vs complete specification, disclosure at fairs (Section 31), renewals, and who owns what when a school is involved. Link to the relevant MYCrave guide.]
Related young innovators
- Trisha Prabhu — ReThink
- Nick D'Aloisio — Summly
- Nilay Kulkarni — Ashioto
Sources
[Editor: list the 2–4 primary sources from the research notes for this entry, with the date checked.]
Working on something like this?
Start with a patentability search, not a filing. Request an expert assessment