A SXSW installation is using deepfake tech to put you inside Malcolm X's speeches, and it might be the first honest use of the technology yet.

The Summary

  • Artist Gabo Arora's "The Great Dictator" installation uses ElevenLabs and Runway AI to insert participants into historical speeches by Malcolm X, Reagan, and NYC council member Zohran Mamdani
  • Instead of fabricating reality, it's using the same tech that powers misinformation to teach people about rhetoric and power
  • You speak a 90-second excerpt, then watch yourself deepfaked into the original footage, delivering the rest in your cloned voice

The Signal

The deepfake conversation has been stuck in a defensive crouch for two years. Every story is about detection, watermarking, or how we're all going to drown in synthetic slop. This project flips that. It's the first real example of using generative AI's power to collapse time and identity for something other than fraud or entertainment.

The mechanics matter here. ElevenLabs clones your voice. Runway stitches your face into archival footage. An AI crowd responds to your cadence and tone in real time. This isn't passive viewing. You're inhabiting the words, feeling the weight of delivery, experiencing what it's like to stand where Malcolm X stood. That's a different kind of learning than reading a transcript or watching a documentary.

Arora's framing is the important part: "humans don't change." He's not saying the tech is neutral. He's saying the real subject is rhetoric and power, which have been tools of persuasion since before writing existed. Deepfakes are just the newest delivery mechanism. The difference is whether you're using them to deceive or to educate.

The speech selection is telling. Malcolm X's "Ballot or the Bullet," Reagan's Berlin Wall demand, and a 2024 city council victory speech. That range shows the installation isn't about ideology. It's about the mechanics of how words move people. You can agree or disagree with any of those speakers, but all three understood how to use a crowd, a moment, and momentum.

This is what generative AI should be doing more of: collapsing barriers to embodied experience. Not replacing humans, not faking reality, but letting people step into contexts they'd otherwise only read about. The same tools that can make a fake CEO confession can also let a high school student in Omaha feel what it's like to deliver a civil rights speech to a hostile room.

The Implication

If you're building with generative AI, this is your permission structure. Stop apologizing for the tech and start showing what it can do when the goal isn't engagement metrics or cost savings. Education, historical preservation, empathy at scale. These are high-value use cases that nobody's chasing because they don't have obvious monetization paths. They should anyway. The companies that figure out how to make tools like this accessible outside art installations will own the narrative on what this tech is actually for.

Source: Fast Company Tech