Home » Latest » Trending TV » Spot on impressions of Al Pacino and Arnold Schwarzenegger by Bill Hader [DeepFake]

Spot on impressions of Al Pacino and Arnold Schwarzenegger by Bill Hader [DeepFake]

4

InfoSearched 67 0 Trending TV

7.5

SaveSavedRemoved 2

A rookie SNL cast member’s talk-show bit just became the internet’s favorite deepfake.

On May 16, 2019, the anonymous YouTube creator known as Ctrl Shift Face dropped a clip titled “Spot on Al Pacino impression by Bill Hader [DeepFake],” and it spread across social media within hours. The video takes old footage from a mid-2000s appearance on “Late Night with Conan O’Brien,” when a then-rookie “Saturday Night Live” cast member named Bill Hader ran through his celebrity impressions for O’Brien’s audience. Ctrl Shift Face didn’t touch a word of audio — instead, the creator layered neural-network face synthesis over the broadcast, morphing Hader’s face into Al Pacino’s whenever Hader slips into that voice.

• Ctrl Shift Face released the deepfake on May 16, 2019, built from archival “Late Night with Conan O’Brien” footage of Hader as a young SNL cast member.
• The video swaps Hader’s face for Al Pacino’s using AI facial-morphing technology, then dissolves back to Hader’s real face when the impression ends — his own voice and mannerisms are left untouched.
• The clip is also referenced online alongside a companion deepfake of Hader’s Arnold Schwarzenegger impression from the same interview footage.

The Source Footage

The raw material here isn’t new — it’s a bit Hader has run through for years on late-night couches, going back to when he was still finding his footing on “Saturday Night Live.” In the Conan interview that Ctrl Shift Face used, Hader drops his own voice and posture and slides into Pacino’s clipped, gravelly cadence, the kind of impression that got him recurring bits on SNL and comparisons to some of the show’s better character actors. That footage is engaging on its own; Hader’s control over Pacino’s specific vocal tics is what made the segment memorable in the first place, similar to how other viral interview clips built their following on genuine, unscripted moments rather than production tricks.

Technical Mechanics of Deepfake Technology

Ctrl Shift Face didn’t fake Hader’s voice or write new dialogue — the entire trick is visual. The software tracks Hader’s facial movements frame by frame and maps a synthesized rendering of Pacino’s face onto them, so every twitch of the eyebrow or curl of the lip lines up with Hader’s own real-time performance. When Hader breaks the impression and laughs as himself, the software dissolves the Pacino overlay away in real time, revealing his actual face again. That seamlessness — no obvious edit points, no glitching around the jawline — is what separated this clip from cruder deepfakes that had circulated online before it.

The face on screen says Al Pacino. The voice, timing, and every ounce of comic instinct still belong entirely to Bill Hader.

Factors Behind Viral Media Success

Part of the appeal is pure novelty — watching a familiar SNL cast member’s face melt into Michael Corleone mid-sentence hits differently than reading about deepfake technology in the abstract. Part of it is that the underlying performance was already good; Hader’s Pacino was praised as one of his sharpest impressions well before anyone thought to run it through AI software. Media outlets and online communities picked it up as both a comedy clip and a case study, the same way outlets have leaned on unscripted celebrity interview moments to demonstrate just how much charisma a good talk-show guest can carry. Here, the added layer of AI-driven video manipulation gave the clip a second life as a talking point about how accessible this kind of software had become by 2019.

Ctrl Shift Face’s video landed at a moment when deepfakes were still mostly a curiosity rather than a widespread concern, and Hader’s clip became one of the go-to examples people pointed to when explaining what the technology could do — not because it was deceptive, but because it was obviously not.