Key Takeaways

  • Ultra-low latency is the next frontier for control. Jean-Baptiste Kempf, co-founder of VLC and FFmpeg, launched Kyber with one specific, audacious goal: achieve 4 milliseconds glass-to-glass latency for real-time machine control.
  • Multimedia is no longer just audio and video. Kempf predicts its definition will expand to include any digital representation for human senses, from odor streams and haptic feedback to full 3D data and brain-computer interfaces.
  • Distance can disappear. Kyber's core mission is to enable “projection of skills or projection of power” by making teleoperation so fluid that the operator feels physically present, regardless of actual distance.
  • The path involves extreme optimization. Achieving 4ms requires pushing networking protocols like QUIC, battling clock drift with precise synchronization, and performing real-time encoding/decoding at unprecedented speeds.

The 4-Millisecond Horizon: Redefining Multimedia

Forget what you thought multimedia meant. Jean-Baptiste Kempf, the mind behind VLC and FFmpeg, isn't just tweaking existing video streams; he's blowing up the definition. His new open-source company, Kyber, aims to conquer an invisible barrier: latency. Kempf's stated goal is starkly specific: “My goal is four milliseconds glass-to-glass latency.” That's faster than your eye can blink, faster than your brain registers a visual change. Most founders are battling 200ms of lag in video calls; Kyber is chasing 4ms for controlling a robot halfway across the world.

This isn't just an engineering challenge; it's a strategic one. Kempf frames it as making “distance disappear because it's either projection of skills or projection of power, right?” Imagine operating a complex machine, a surgical robot, or a drone in a dangerous environment, with the same responsiveness as if you were physically there. That kind of real-time connection isn't about better video quality; it's about enabling entirely new forms of remote control and presence. Kyber achieves this by optimizing every layer: ultra-efficient networking using protocols like QUIC, aggressive real-time encoding, and meticulous synchronization to neutralize clock drift.

Beyond Pixels: Your Senses Are the New Streams

For Kempf, the expanding definition of multimedia is the bedrock of this future. He puts it plainly: “In order to answer that question, we need to answer the definition of what is multimedia. And multimedia is a digital representation of several streams for the human senses. And we will do that, right?” This isn't theoretical. He points to existing capabilities and future integrations:

“There is now a way to not have a mic, but have an odor sensor- and a diffuser of odors. It will get into FFmpeg.”

And it goes even further: “It's not stopping. And yes, at some point it will manage 3D data inside VLC and FFmpeg, right? It's obvious.” This implies a future where VR/XR isn't just visual and auditory, but fully immersive, integrating haptic feedback, environmental scents, and even brain-computer interfaces as just another 'stream' that needs to be compressed, transmitted, and rendered with minimal delay. Your senses become the data points, and systems like Kyber are building the pipes to carry them.

What to Do With This

Stop thinking about your product's data streams in terms of just video or audio. This week, map out every single 'human sensory input' and 'machine control output' relevant to your product. For each, identify its current latency and, more importantly, what a 10x reduction (e.g., from 40ms to 4ms) could unlock. Could drastically lower latency enable a new remote service, a more immersive user experience, or even a completely new business model where distance is no longer a factor?