If you have watched a video online in the past two decades, you have used FFmpeg or VLC. These two open-source projects underpin almost every piece of video software you touch. Yet the people who build them rarely sit down for a public conversation. That changed with episode 496 of the Lex Fridman podcast, where host Lex Fridman talked with Jean-Baptiste Kempf, lead developer of VLC and president of VideoLAN, and Kieran Kunhya, a longtime FFmpeg contributor, codec engineer, and the person behind the now-infamous FFmpeg account on X.

The conversation runs four hours and covers everything from the weirdest file formats VLC can open to the internal politics of the FFmpeg project, the use of hand-written assembly code, and the ongoing push for the next-generation video codec AV2.

Who they are and why it matters

FFmpeg is a command-line multimedia framework that can decode, encode, transcode, mux, demux, stream, filter, and play almost anything that moves on a screen. It is the engine inside countless products: YouTube transcoding pipelines, OBS Studio, HandBrake, popular video editors, and even some hardware devices. VLC is the media player that opens nearly any file you throw at it—including damaged or incomplete files. Kempf has been its lead developer for years, steering a project that is downloaded hundreds of millions of times.

Kunhya is a core FFmpeg developer who works on low-level codec implementation. He also commandeered the official FFmpeg account on X (formerly Twitter) and turned it into a mix of technical threads, commentary on open-source drama, and occasionally blunt takes on the state of video engineering.

What they covered in the episode

The conversation starts with the weirdest things VLC can open—damaged files, incomplete streams, raw bitstreams that other players reject. From there it moves into the mechanics of video playback: how a video file is parsed, how frames are decoded and presented in sync with audio.

The middle third of the episode digs into FFmpeg itself. Kempf and Kunhya explain the difference between a codec (the algorithm that compresses video) and a container (the file format that holds the streams). They also discuss the project’s history: FFmpeg was created by Fabrice Bellard in 2000, and has since accumulated a sprawling codebase in C with hand-optimized assembly for critical inner loops. Kunhya explains why some routines are still written in assembly—performance that cannot yet be matched by compilers for certain SIMD operations.

The drama: Turning down millions and the Google incident

One of the most striking sections is Kempf’s description of turning down millions of dollars to keep VLC ad-free. According to the discussion, large companies have offered significant sums in exchange for integrating advertising into VLC. Kempf and the VideoLAN team have refused every time, citing the project’s commitment to being a pure, neutral tool.

The episode also covers the widely discussed “FFmpeg and Google drama.” Kunhya and Kempf recount a situation where a major corporation (apparently Google) used FFmpeg in a way that violated the project’s licensing terms, leading to a public dispute and a rearrangement of how FFmpeg handles contributions from large companies.

Technical depth: Low-latency streaming and codec patent politics

For the technically inclined, the episode includes a deep discussion of ultra-low-latency streaming protocols, x264 (the popular H.264 encoder), and the basics of video compression: intra-frame vs inter-frame prediction, motion vectors, and bitrate control. Kempf explains that much of the progress in streaming today—think live gaming or responsive conferencing—depends on reducing the latency introduced by encoding pipelines.

Toward the end, the conversation turns to the patent landscape. Video codecs are heavily patented, and the next-generation standard AV2 (Alliance for Open Media project) aims to be royalty-free. The hosts discuss the challenges of implementing new codecs in open-source projects while avoiding patent litigation, and the role of groups like the MPEG LA in shaping what codecs ship.

The darker side: Backdoors, fake VLC, and burnout

Kempf reveals an unusual story about the CIA reportedly distributing a fake version of VLC as part of a surveillance operation. The anecdote, sourced from public disclosures, highlights how widely used open-source tools become targets for intelligence agencies.

The episode also addresses open-source burnout. Both Kempf and Kunhya talk about the difficulty of maintaining projects under constant pressure from users, companies that demand features without contributing code, and the endless stream of bug reports and entitlement.

Future of FFmpeg and VLC

The episode ends with a look ahead: What’s next for video on the internet? Kempf discusses VLC’s experimental work with hardware decoding and AV1 support. Kunhya talks about the ongoing rewrite of FFmpeg’s filter infrastructure and the possibility of a more modular future for the project. Both express cautious optimism about the adoption of AV2 as a truly free codec.

Why this episode matters for anyone who touches video

Most people open a video file and expect it to just work. The unseen miracle behind that expectation is two volunteer-run projects—FFmpeg and VLC—that have been quietly maintained for over twenty years. The Lex Fridman episode gives those projects and their developers a rare spotlight, combining technical explanations with the human stories of building tools that billions rely on every day.

Whether you are a software engineer wanting to understand the internals of video decoding, or a regular user curious about why VLC can open a corrupted .avi from 2004, the conversation offers a valuable—and often entertaining—look behind the curtain.