Top 5 in AI

Guides

How to Make the 'All-Loss Tiny Desk' AI Video: Losing College Football Coaches Face-Swapped Into Charlie Wilson's NPR Set, Who Made It, and What It Takes

By the Top5Apps editorial team · Published October 20, 2026 · Updated October 20, 2026 · 10 min read

Share
AI face-swap frame: Steve Sarkisian in a white Texas pullover, Kirby Smart in a Georgia cap, and a coach in a red Adidas jacket singing into Tiny Desk headset mics
The AI version: Sarkisian, Smart, and a coach in Indiana red singing Charlie Wilson's backing parts at the Tiny Desk. Frame from The Burner's video, shown editorially.

Short answer: the video is a multi-person face swap laid over about nineteen seconds of Charlie Wilson's real NPR Tiny Desk concert from June 5, 2023, specifically the closing number, the Gap Band's 'Outstanding.' Everything you hear and every body movement is NPR's footage; only the faces are the coaches who lost on Saturday, October 17, 2026. It was made and posted by a TikTok account called The Burner (@theburnerfromhe11), who labels it 'This is an AI parody!' and names no tool. To make one yourself you need the source clip, one clean reference photo per coach, a video face-swap tool that can handle several people at once (CapCut, DeepSwap, Remaker, or the open-source FaceFusion and Deep-Live-Cam), and about an hour. The method below is reconstructed from what those tools do, not from the creator, who has said nothing about process. Every claim in this guide is linked; the ones we couldn't verify are listed near the end.

Why this weekend produced it

Saturday, October 17 was the strangest top-of-the-poll day in college football history. Per CBS Sports, five of the top seven teams in the AP Top 25 lost on the same day for the first time since the poll began: No. 1 Georgia fell 27–24 to unranked Auburn on a field goal as time expired; No. 2 Notre Dame lost 28–24 at No. 8 BYU; No. 5 Indiana's 22-game winning streak ended 19–15 against No. 4 Ohio State; No. 6 Alabama lost 26–10 at No. 14 Tennessee; and No. 7 Texas lost 27–24 at home to unranked Florida. Scores are from ESPN's final box scores. That is Kirby Smart, Marcus Freeman, Curt Cignetti, Kalen DeBoer, and Steve Sarkisian, the five coaches the creator tagged, all losing within about seven hours of each other. A video of them consoling each other with a funk standard about being outstanding wrote itself; someone just had to render it.

The chain of posts, verified where we could. The Burner posted it first to TikTok at roughly 7:55 pm ET on Sunday, October 18 (the time is derived from the video ID; TikTok's page only says '1d ago'), with the caption 'College Football's Week 8 All-Loss Tiny Desk Concert 😭🎤 #alabama #collegefootball #indiana #georgia #texas This is an AI parody! Original Video is Tiny Desk: Charlie Wilson. Original Song is Outstanding by the Gap Band.' TikTok shows its 'Contains AI-generated media' label on it. About six minutes later the same video went up as an Instagram reel with the same caption, this time tagging #notredame instead of #collegefootball. As of Monday morning the TikTok sat at 1.7 million plays, 239,800 likes, and 102,700 shares; the reel at 83,300 likes. The account is new and small: the TikTok profile has one public video and 1,408 followers, the Instagram one post and 1,349 followers, and the X account (@burnerfromhe11) was created October 5 and has thirteen followers. Its bio: 'Just posting what I think will be funny for the week. Yes, I made these videos.' The X post that made it explode, from @propjoesays ('string,' a Texas A&M poster with 37,500 followers) around 10:10 pm ET with the caption 'what the hell is happening,' is the one link in the chain we could not retrieve; X blocks every mirror we use, and the post isn't indexed. We're citing it as reported, with the packet's figures of roughly 12,000 likes and 2.7 million views, and will replace that with verified numbers once we have the post ID.

The real Tiny Desk, and what the swap keeps

The source, official NPR Music upload, June 5, 2023, 29:38 long, 6.87 million views as of October 20, 2026. Cued to 23:51, where the excerpt the parody uses begins; 'Outstanding' closes the set.

NPR published Charlie Wilson's Tiny Desk on June 5, 2023 as part of its Black Music Month series, produced by Bobby Carter. The set runs 'Charlie, Last Name Wilson,' 'Computer Love,' 'There Goes My Baby,' 'I'm Blessed,' 'Yearning for Your Love,' 'Sweet Yamz,' and then 'Outstanding,' the Gap Band single Raymond Calhoun wrote, released in November 1982 on Total Experience Records and a No. 1 R&B hit in February 1983. NPR's credits list eleven musicians behind Wilson: five backing vocalists (Jor'el Quinn, Leah Thompson, Gia Wyre, Dodie, Angel Hines), music director Cordell Walton and two more on keys, Alex Williams on guitar, bass from Lawrence Washington and EQ, Rain Man on drums, and Nick Wilczynski on saxophone. That saxophone slot is the one the packet and commenters say Marcus Freeman fills in the parody; we couldn't view the full video without a login, so we're reporting that as what viewers describe rather than something we confirmed frame by frame.

Charlie Wilson in a blue beanie, gold aviators, and a baroque-print jacket singing at the NPR Tiny Desk, saxophonist behind him
The original: Charlie Wilson closing with 'Outstanding,' Nick Wilczynski on sax behind him. NPR Music frame, shown editorially.

Source clip

The 14-second NPR excerpt the parody is built on, 1280×720 with audio, so you can try the swap yourself.

Download the clip (2.5 MB) ↓

It starts at 23:51 of NPR's upload, which we confirmed by matching the clip's audio against the concert. The footage is NPR's; it's offered here as a short excerpt for commentary and instruction, with the full performance embedded above. If you post what you make, credit NPR, Charlie Wilson, and the Gap Band in the caption the way the original creator did.

  • What stays: the audio, every note of it, including the crowd of NPR staffers; the camera moves; the Tiny Desk shelves; every performer's body, hands, instruments, and lip movement. In the frame at the top, Wilson's blue beanie, gold aviators, and baroque-print jacket are gone, replaced by a Texas pullover, a Georgia cap, and an Adidas jacket, which tells you the swap is doing heads and some wardrobe, not just faces.
  • What changes: identities only. The singers' faces become coaches' faces, with the coaches' expressions driven by the original performers' expressions. That is why the comments single out how natural it looks: nobody generated a coach singing, they borrowed a professional's performance and re-skinned it.
  • What a human did: picked a nineteen-second slice of a 29-minute concert where the backing singers are on camera and emoting, so the swap had faces to work with; chose coaches whose losses were fresh; and matched who-goes-where with some care (the sax player's build and glasses are a plausible Cignetti or Freeman stand-in). Those choices are the comedy. The software is the plumbing.

Method (reconstructed — the maker said only 'This is an AI parody!')

The Burner has not named a tool, described a workflow, or answered the 'who made this' questions in the comments beyond the bio line. What follows is how a multi-person face swap on existing footage is done with the tools available in October 2026. It is a reconstruction; a different tool could produce an identical result.

  • 1. Get the source clip. The 14-second excerpt above is the one the parody uses, and it's ready to drop into a swap tool. If you want more of the song, the embed is cued to 23:51 of NPR's upload; screen-record from there at the highest resolution you can (the master is 1080p) and trim to where the backing singers face the camera. Keep the audio untouched; it is the whole joke.
  • 2. Collect one reference face per coach. A front-facing, evenly lit, neutral-to-smiling photo at least 512 pixels across the face: a press-conference frame or a sideline shot. Match the angle of the performer you're replacing where you can. Headset-and-visor shots are bad references because the headset occludes the face.
  • 3. Pick a tool that handles several faces in one video. Consumer web tools: DeepSwap advertises refacing 'up to 6 faces simultaneously in the same video clip' and bills by credit (one credit per fifteen seconds of video, pricing in credits rather than dollars on its page); Remaker has a dedicated 'Video Multiple Face Swap' tool, 500 MB or 30 minutes max, at two credits per second, though reviews conflict on whether a free tier still exists; CapCut has a video face-swap tool inside its web editor, with its AI features generally gated to the roughly $20-a-month Pro plan. Open source, run locally on a decent GPU: FaceFusion (version 3.9.1, released September 30, 2026, 30,000 GitHub stars) and Deep-Live-Cam (2.7.5, 97,000 stars), whose Face Mapping mode is built for exactly this: 'use different faces on multiple subjects simultaneously.'
  • 4. Map faces to performers. In the multi-face tools you load the video, the tool detects each distinct face, and you assign a reference to each one. Assign all of them, including performers who only appear for a second, or the swap will flicker between a coach and the original musician. If your tool only does one face at a time, run it once per performer, each time feeding the previous output back in as the new source.
  • 5. Run a face-enhancer pass. Swapped faces come out slightly soft and sometimes a shade off from the scene's lighting. FaceFusion's and DeepSwap's built-in enhancers (GFPGAN or CodeFormer under the hood) sharpen and re-light them. Don't push it to maximum; over-enhanced faces look waxy and give the game away.
  • 6. Check the hard frames. Head turns, hands crossing faces, and the sax mouthpiece are where swaps break. Scrub those moments. If one is bad, either cut around it or re-run that performer with a reference photo at a closer angle.
  • 7. Export, label, credit. Re-mux with the original audio if your tool dropped it, export at the source resolution, and do what the creator did: say it is an AI parody in the caption and name the source performance and song. TikTok's AI label is applied automatically when the platform detects it, but the caption is yours.

What this is not. It is not generated video. None of the models in our AI video generators ranking made these frames; Veo, Kling, and Runway would give you a new performance with new motion, which is a different, harder, and less funny project. The face-swap category is older, cheaper, runs on a laptop, and is better at exactly this: preserving a real performance while changing who is giving it. If you want to go the generation route instead, the image generators are where you'd build the coach reference frames first; our Usual Suspects line-up guide covers a five-person scene swap done that way.

What the packets got wrong, and what we couldn't confirm

  • It wasn't Week 8 or Week 9. The creator's caption says Week 8 and one widely shared X post says Week 9. ESPN's scoreboard and CBS both label October 17 as Week 7 of the 2026 season (Week 0 was August 29, Week 1 began September 3). Funny videos don't need calendars; guides do.
  • It wasn't posted October 17. The earliest copy is the TikTok at about 7:55 pm ET on October 18, the day after the games, with Instagram six minutes behind. Nothing from October 17 exists on any of the creator's accounts.
  • The X numbers are unverified. Every mirror we use to read X without logging in is down or blocked this month, and the @propjoesays post isn't indexed. The '12,000 likes, 2.7 million views' figures are from the packet. Same for the Kameo Williams and Melissa posts the packet cites as thirty minutes later.
  • Who plays sax. Viewers say Marcus Freeman. We could not watch the full clip without an account and won't assert a face we haven't seen.
  • The tool. Unnamed. Any article telling you 'it was made with CapCut' is guessing.
  • 'First viral hit.' The account is fifteen days old with one public video, so this is its only hit that we can see; the bio promises weekly content. Two more videos exist in TikTok's count but aren't publicly visible.

Rights: NPR's footage, the Gap Band's song, five public figures

Three layers. The footage is NPR's, produced by its staff in its office; a nineteen-second excerpt reworked into a parody of the coaches, not of the concert, is a plausible fair-use argument but not a settled one, and NPR could ask the platforms to pull it on copyright grounds without going near the AI question. The recording is Charlie Wilson's performance of a composition by Raymond Calhoun; the sync use of the audio is the most exposed layer, and it is the same exposure every meme that keeps original audio carries. The faces are five head coaches at public universities, public figures being parodied for their public performance in a game; that is the easiest layer, and the tradition of ribbing coaches on a bad Saturday is older than the forward pass. What changes the calculus is scale and intent: the video is labeled as parody, keeps NPR's branding in frame, credits the source and the song in the caption, and sells nothing. As of October 20 neither NPR, nor any of the schools, nor Wilson's camp has reacted publicly, and both posts are up with their AI labels intact. If you make your own, keep every one of those habits.

Our read

This is the best kind of AI meme: a real cultural artifact, a real sporting fact, and a joke that only works because both are real. The craft is in the selection, not the software. Picking a Tiny Desk, where the camera stays on the backing singers' faces, instead of a stadium clip; picking 'Outstanding' for five men who were not; posting on Sunday night when every one of those fan bases was still stewing. A multi-face swap tool did the rest in minutes. For creators the lesson is the same one the Usual Suspects and Rumpelstiltskin guides taught: find footage with a performance worth stealing, label it, credit it, and let the week's news pick the cast. For readers trying to tell what's real, note what gave this away even before the label: coaches don't harmonize, and a Georgia cap at the Tiny Desk is a tell. The faces will keep getting better. The incongruity won't go anywhere.

Where these apps rank