AI-Driven Automation for Corporate Meeting Video Capture

A leading corporate meeting media company, challenged us to help revolutionize their business model with AI-driven technology, transforming camera operations with intelligent speaker tracking and dynamic video management, while overcoming tight deadlines and complex scalability requirements.
close-up-company-team-working
A client approached us with an ambitious vision to revolutionize corporate event spaces using AI-driven technology. Their goal was to replace human-operated cameras with an intelligent system capable of automatically focusing on speakers and delivering seamless, professional-grade video streams. The system would utilize advanced AI models to identify and track active speakers in real time and integrate seamlessly into high-end event settings.
No data was found
Witekio services highlight
  • Hardware selection and evaluation
  • Software redesign
  • Algorithm development
  • Scalability testing

Witekio in action

Challenges

  • Tight deadlines: They aimed to complete the project in under two months.
  • Limited internal expertise: The client didn’t have the internal knowledge on how to build industrial grade software and relied on prototypes built using basic Python and Google tools.
  • Untested end-results: The existing prototype was functional for limited scenarios but lacked scalability and production readiness.

Businessman talking with remote team planning management strategy

Solution by Witekio

Our team began by conducting an in-depth analysis of the client’s prototype and technical requirements. This involved assessing the existing solution’s capabilities and limitations when deployed on NVIDIA AI hardware.

  • Hardware Selection and Evaluation: We determined that the client’s initial hardware choices, while promising, required optimization for real-world performance. NVIDIA’s tools were selected for their AI processing power, but further testing revealed gaps in the hardware’s compatibility with the envisioned application.

 

  • Software Redesign: The original prototype’s reliance on Python and generic tools was not suitable for scaling. We migrated the solution to a robust C++ stack tailored for NVIDIA’s ecosystem. This step significantly improved performance and allowed for deeper integration with advanced AI models.

Software-factory-3-ressource-visuel

Solution by Witekio

  • Algorithm Development: Our team developed custom algorithms for:
    • Refining a speaker detection algorithm using facial feature analysis (e.g., lip movement and gaze tracking) developed by the client's internal team.
    • Camera management logic for dynamic focus adjustment.
    • Multi-camera input merging to create seamless video streams.

 

  • Scalability Testing: Extensive testing with pan-tilt-zoom cameras revealed further limitations in off-the-shelf AI models. Collaboration with NVIDIA’s R&D team confirmed that additional customizations were necessary to meet the client’s requirements.

Results

While the technical foundation for the project was successfully laid, it became evident that the client’s desired timeline and budget were insufficient for full implementation. Key outcomes included: 

  • Detailed Feasibility Study: We delivered a comprehensive roadmap outlining the steps, costs, and timelines required to bring the solution to market. The analysis highlighted the need for a one-year development cycle to achieve production-ready AI models.
  • Prototype Refinement: The re-engineered prototype demonstrated improved functionality, showcasing the potential of AI-driven camera systems in controlled environments.
  • Knowledge Sharing: Documentation of lessons learned and best practices was compiled, forming the basis for future projects and contributing to internal expertise.

We transform your device vision into reality

We support your teams in designing, building, and running innovative products, from embedded software to application development
flag_line

4 Countries

4 countries

Witekio is certified ISO 27001

ISO 27001 certified

ISO 27001 certified

Avnet_logo

fortune 500 owned

fortune 500 owned

Contact us