Inspiration
Videos contain massive amounts of knowledge, but most of it is trapped in timelines. We wanted to make video content as searchable and accessible as documents by helping users find answers hidden in speech, visuals, slides, and gestures.
What it does
Meridian transforms videos into searchable knowledge. It analyzes spoken words, on-screen text, and visual scenes, allowing users to ask questions and jump directly to the exact moment that contains the answer.
How we built it
We built Meridian using three parallel pipelines: speech transcription with timestamps, OCR for extracting text from frames, and visual understanding for describing scenes. An AI reasoning agent combines these sources to retrieve the most relevant video moment.
Challenges we ran into
The biggest challenge was understanding that video information is multimodal. Connecting spoken context, visual details, and text while maintaining accurate timestamps required careful processing and retrieval design.
Accomplishments that we're proud of
We created a system that can uncover information that traditional video search misses—like content written on whiteboards, shown in slides, or communicated visually.
What we learned
We learned that making video searchable requires more than transcription. True understanding comes from combining language, vision, and reasoning together.
What's next for Meridian
We plan to improve accuracy, support larger video libraries, add better collaboration features, and make Meridian a complete knowledge search engine for organizations.
Log in or sign up for Devpost to join the conversation.