Top 5 in AI

Guides

How to Make the Anakin-and-Stormtroopers 'Gangsta's Paradise' Star Wars AI Video: What the Clip Actually Is (Not a Face Swap), Who Posted It, the Two Ways to Build a Scene That Doesn't Exist, and the Rights Problem Nobody in the Replies Mentions

By the Top5Apps editorial team · Published November 6, 2026 · Updated November 6, 2026 · 8 min read

Share

Short answer: this isn't a face swap on a movie scene, because the scene doesn't exist. It's a fully generated 30 seconds in which a prequel-era Anakin sings a 1995 rap hit into a wireless mic in an original-trilogy hangar, a line of stormtroopers sways behind him, and Obi-Wan takes the mic for the hook. The post that made it travel, 'Star Wars ai is getting better and better 😂' from the aggregator account @dripbitsnft on November 5, had 4,536 likes, 568 reposts and 519,849 views when we checked, credits nobody, and no reply we could reach names a tool. So the method below is reconstructed from what the frames show and what the current tools can do, and it is labeled that way.

AI-generated frame: a man resembling Hayden Christensen as Anakin Skywalker in dark Jedi robes sings into a handheld microphone in a white hangar, with a row of stormtroopers holding blasters behind him and a man resembling Ewan McGregor as Obi-Wan Kenobi at right
The clip as it circulated, shown editorially. Anakin on the mic, Obi-Wan waiting, troopers in formation. Note the hangar: it is Death Star white with Imperial stormtroopers, not the clone-era sets the costumes come from.

What the clip actually is

Thirty seconds, 960×720, 30 frames a second, a single continuous performance with a few cuts. The staging is a revue: lead singer front and center with a modern handheld mic, backup line behind, everyone in the same key light. Anakin wears his Revenge of the Sith costume and Hayden Christensen's face; the troopers are Imperial stormtroopers from the original trilogy, E-11 blasters and all, not clones; the room is a white-walled hangar with a starfield outside, the kind the first film shot on. Around the 27-second mark Obi-Wan, bearded and in Ewan McGregor's likeness, steps in from the left and takes the mic for the hook while Anakin stands back. The troopers' choreography is a hand-to-chest sway, the kind of thing doo-wop backing singers do, and it holds across the trooper-line cut, which is the detail that made people argue about whether it was real.

AI-generated frame: a man resembling Ewan McGregor as Obi-Wan Kenobi at lower left sings into a handheld microphone while a man resembling Hayden Christensen as Anakin stands behind him and a row of stormtroopers sways with hands clasped
The handoff at 27 seconds, shown editorially. Obi-Wan takes the hook; the troopers keep the sway.

Two things people are getting wrong. First, the 'deepfake over real footage' description: there is no Star Wars shot with these two characters, a microphone and Imperial stormtroopers in a Death Star hangar, so nothing was swapped onto film. What is reproduced is the two actors' likenesses, which the model renders well enough to be recognizable at phone size and to wobble under scrutiny (watch Anakin's hairline in the push-ins). Second, the song: our editor identifies it as Coolio's 'Gangsta's Paradise', the 1995 single with L.V. that interpolates Stevie Wonder's 'Pastime Paradise'. We could not fingerprint the audio ourselves and no text source names it, so take that as an ear, not a record. We quote no lyrics here, and the rights section explains why the song is the expensive part of this clip, not the Jedi.

Dripbits@dripbitsnft

Star Wars ai is getting better and better 😂

View original post ↗

The account is a reposter, not the maker: its two earlier Star Wars AI hits both carry 'Who did this?' captions, including the pinned Lil Vader 'Mace fell out of a Windu' clip at 2.96 million views, which originated with the TikTok creator Ai Ruins Everything and was reposted by Elon Musk in October at 4.5 million views, and the 'Reggae Star Wars' series by the TikTok creator @aiamajedi at 1.08 million. Star Wars characters performing music is a 2026 genre now, and this clip is the first of the three to put the prequel actors' faces on it rather than a helmet. The only realism 'debate' visible without logging in is one reply: 'to anyone who thinks this is AI. this video is 100% real. i asked claude and it agreed.'

What stays, what changes

  • Stays from the source material: the two characters' costumes and faces, the stormtrooper armor and blasters, the Imperial hangar look, and the track, untouched under the whole clip.
  • Changes: everything else. The scene, the blocking, the microphone, the choreography and the camera moves are new. This is the opposite of the Stromae and Yung Lean 'Storm' swaps, where a real video's timing and cuts carry the joke and only the faces move. Here the joke is that the staging is a 1960s TV revue and the cast is the Empire.
  • What the 'Yung Lean edits' comparison gets right and wrong: right that those copies normalized putting famous faces on a famous performance; wrong on the mechanism. Per Stereogum, most 'Storm' copies were made with Higgsfield's Genjutsu, which swaps a character into an existing video. This clip has no existing video under it.

How to make it (reconstructed; nobody published this one)

The two routes. The one-pass route matches the clip's look (one lighting setup, consistent cast across cuts); the driving-video route is how you'd get the mic timing exact.
  • 1. Make the audio first. Trim the track to the 30 seconds you want, with the hook landing in the last five so the handoff has somewhere to go. Every tool below takes an audio file; none of them will fetch the song for you.
  • 2. Build a reference set, not a prompt. Five stills do the work: a clean frontal of Anakin in the dark Jedi robes, one of Obi-Wan, one stormtrooper, one wide of a white hangar with a starfield, one handheld mic. Still images of the characters are your choice to make; see rights before you make it.
  • 3. Route A, one pass: Seedance 2.5 takes up to 30 images, 10 video clips and 10 audio clips and generates 'high-quality, 30-second audio-video clips in a single pass'; its own sample prompt is 'All subjects lip-sync while singing.' Kling 4.0 does the same with Omni Reference ('up to 15 reference assets in a single generation, combining images, videos, and subjects'), 30-second generations and lip-sync that keeps 'voices, dialogue, and performances naturally synchronized'; Kling 4.0 Flash reached Ultra Yearly subscribers first and the full model shipped October 15.
  • 4. Route B, performance transfer: film yourself singing it with a mic, locked camera, front light. Runway Act-Two 'only requires a driving performance video and reference character' and tracks 'head, face, body and hand', 30 seconds maximum, 5 credits a second (Runway retired its separate Lip Sync tool in May). That gives you the Anakin take with your timing. Generate the trooper line as a separate Route A shot and cut between them, which is roughly what the clip's trooper-line cut looks like anyway.
  • 5. Prompt (ours, not the maker's):
A 1960s television revue staged in a white Imperial hangar with a starfield visible through the open bay. @Image1 (Anakin, dark Jedi robes) stands front and center singing into a handheld wireless microphone, performing to @Audio1 with exact lip-sync and expressive gestures. Behind him a straight line of eight stormtroopers (@Image3) sway in unison, hands to chest, holding blasters, as backing singers. Hard key light, slight push-in on the singer, one cut to the trooper line at the midpoint. In the final five seconds @Image2 (Obi-Wan, light robes, beard) walks in from the left, takes the microphone and sings the hook while @Image1 steps back. Keep the costumes and armor exactly as in the images. No subtitles.
  • 6. Expect to rerun it. The failure modes are the usual ones: a trooper who grows a second blaster, a hairline that drifts, lips that go soft on fast lines. Four to six generations is normal; keep the one where the handoff lands on the hook.
  • 7. Export at 4:3 if you want it to look like this one. The circulating clip is 960×720, which is why it fills a phone's width with letterboxing; most tools default to 16:9 or 9:16.
  • 8. Label it. 'AI' in the caption, not just a laughing emoji. Then read on.

Rights: two living actors, one dead rapper's song, and the most litigious studio in the business

This clip stacks three claims. The song is the clearest: 'Gangsta's Paradise' is written by Coolio, L.V. and Doug Rasheed with Stevie Wonder credited for the 'Pastime Paradise' interpolation; its master belongs to Reservoir Media, which bought Tommy Boy's catalog in 2021, and its publishing sits with Sony Music Publishing's Jobete/Black Bull (Wonder's share) and Warner Chappell. Coolio died in September 2022; his estate and those publishers are who a platform's Content ID pays or who files the claim. Thirty seconds of the master under a parody is not a fair-use safe harbor, and 'Amish Paradise' is the famous reminder that even Weird Al's parody of this song caused a dispute.

The actors are the newer problem. Hayden Christensen's and Ewan McGregor's faces are what sells the clip, and every major tool's policy touches it: HeyGen bans content that would 'represent real individuals, including celebrities or public figures, without their explicit consent'; Runway's usage policy prohibits 'use of an image, video, or audio of another person without their permission'; Google's prohibited-use policy bars impersonating 'an individual (living or dead) without explicit disclosure, in order to deceive' and runs a likeness-complaint process; Hedra bans uploads 'containing individuals other than yourself'; Kling's community guidelines ask you to respect 'rights of reputation, likeness, privacy'. None names an actor, and all of them leave 'obvious parody with a disclosure' to the platform's judgment. When Disney licensed its characters to OpenAI's Sora last December, the deal pointedly did 'not include any talent likenesses or voices', which tells you where the studio draws the line even when it is being paid.

And Disney owns the room. Lucasfilm has made no 2026 statement about fan-made AI Star Wars videos that we could find, but Disney sent ByteDance a cease-and-desist over Seedance 2.0 in February, citing Darth Vader among the characters and calling the model's output a 'virtual smash-and-grab', and sent Google one over Veo in December, after which Google removed videos. The practical reading: a 30-second parody on a personal account is the kind of thing studios have mostly let ride; the same clip monetized, or used to sell anything, is not. If you make one, don't run ads on it, don't put a brand in it, and don't be surprised if the tool refuses the faces.

Our read

The interesting thing here isn't the faces. It's that nothing in the frame existed before, and it still reads as a scene. The 'Storm' copies and the Stromae swaps borrowed a real director's timing; this one had to invent a stage, a lighting setup, a dance for eight extras and a camera, and keep all of it consistent across cuts for 30 seconds. That is what the one-pass models learned to do this year, and it is why 'Star Wars AI is getting better' landed as a caption: the improvement people are noticing is not the deepfake, it's the staging. Our video rankings score consistency across cuts for exactly this reason. The best-performing joke in the clip is also the oldest trick in music video: put the straight man on the hook. Obi-Wan taking the mic is the shot, and whoever made this knew it.

What we could not verify

  • Who made it and with what. The post credits nobody; the four replies visible without logging in name no tool; the 75 quote posts are behind a login. Both routes above are our reconstruction.
  • The song. Our editor's identification; we did not run an audio fingerprint and no text source names it.
  • Any live performance the choreography is traced from. We found none; the sway and the handoff are generic revue staging.
  • Which Runway plans include Act-Two. The pricing page doesn't say; third-party guides say Standard and up.
  • The 4:3 output's origin. Whether it came from a tool default or a crop is unknown.

Where these apps rank