cd /news/generative-ai/no-scrubs-ai-video-from-one-group-ph… · home › topics › generative-ai › article
[ARTICLE · art-171646] src=dev.to ↗ pub= topic=generative-ai verified=true sentiment=· neutral

No Scrubs AI Video From One Group Photo: Modelling the Left-Middle-Right Casting Rules

A developer reverse-engineered the casting logic behind the "No Scrubs" AI video trend, modeling how a single group photo is mapped onto the template's three unequal slots. The documented rules reduce to sorting detected faces by area, keeping the three largest, then sorting left to right, so the middle position gets five of six shots while the right slot appears only in the stage row. The developer recommends cropping to three people or switching to the per-person photo mode when a specific trio is wanted.

key takeaway

For a three-slot template, cast by sorting detected faces by bounding-box area, keeping the top three, then sorting by horizontal position, since the largest faces win entry and left-to-right order assigns roles.

by read4 min views7 publishedNov 18, 2026

The "No Scrubs" AI trend puts three new people into TLC's 1999 music video: out of the glowing white corridor, into the camera for close-ups, then a row of three on the stage with the big TLC letters behind them. Spider-Men, sitcom grandmas, a reality-show cast. The joke is always who the three are.

What makes it interesting to a developer is that a trio template has a casting problem. Three positions are not equal, and if you feed in a single group photo, something has to decide who goes where. The No Scrubs AI video generator documents its rules pretty precisely, so I turned them into a small function. Writing them down this way made it obvious how to arrange a group photo before rendering.

The template is based on the version @herokobaa posted on October 20, 2026. It has six shots, always in the same order:

Counting screen time from the site's breakdown:

Position Shots Notes
Middle 5 of 6 The only required photo
Left close-up + stage solo + row The "spotlight" friend
Right row on the stage only No close-up of its own

So the middle slot needs the sharpest photo, the left slot needs a face that holds up in a close-up, and the right slot is the most forgiving.

Two input modes exist: A photo each (three boxes: You, Friend on the left, Friend on the right) or One group photo (up to three people). Here is what the site says happens, expressed as JavaScript. This is my model of the documented behaviour, not their implementation.

// faces: [{ x: centerX, area: boxArea, id }]
function castFromGroupPhoto(faces) {
  // "With more than three people, the three biggest faces go in."
  const picked = [...faces]
    .sort((a, b) => b.area - a.area)
    .slice(0, 3)
    // "Left to right in the photo is left, middle and right on the stage."
    .sort((a, b) => a.x - b.x);

  if (picked.length === 3) {
    return { left: picked[0], middle: picked[1], right: picked[2] };
  }
  // Fewer people: an empty slot keeps the original dancer.
  // (How 1-2 faces are assigned in group mode isn't documented,
  //  so for anything but a full trio use "A photo each".)
  return null;
}

function castFromPhotoEach({ you, left, right }) {
  if (!you) throw new Error("The middle box is the only required photo");
  return {
    middle: you,
    // Leave left empty and *you* also take the left close-up + solo;
    // the original dancer keeps the left of the row.
    left: left ?? "original dancer (your face in the close-up and solo)",
    right: right ?? "original dancer",
  };
}

Once you see the rules as sort by area → take 3 → sort by x, a few practical tips fall out:

1. Size matters more than position for who gets in. If four people are in the shot, the person furthest from the camera is the one dropped. If you want a specific trio, crop the photo to three people first.

2. Then position decides the role. Whoever stands in the middle of the picture walks out of the corridor and gets five of six shots. Put your most photogenic or most committed friend there.

3. The left of the picture is the close-up seat. That face appears big on screen, so it needs to be sharp and front-on.

4. When in doubt, switch modes. "A photo each" removes all ambiguity: you choose who sits in which box. The site suggests exactly that if you want to choose for yourself.

From the site's own guidance, each face should be:

One detail worth knowing: the original dancers wear a helmet with goggles, a beret and a scarf. The site says the output is made to show each whole face, hair and clothes from your photos instead. If something from the original costume survives, their advice is to try a clearer photo.

The flow itself is short: add photos, tick "I have the right to use these photos.", press Make my No Scrubs AI video. Nothing to type. Most renders take 5 to 10 minutes and you get an email when it is ready.

Output facts worth planning around:

Packs are one-time: $9.90 for one Standard video, $19.90 for two HD, $39.90 for five HD. I did not see a free tier on the site.

The terms only allow photos of you or adults who have agreed, and they exclude anyone under 18 and public figures. With a trio template you are casting two other people, so ask first, then tag them when you post (with the AI-generated label on).

Writing the casting rules as code takes the guesswork out of a group render: crop to three, put your star in the middle, put your sharpest close-up face on the left. Then let the template do the choreography.

── more in #generative-ai 4 stories · sorted by recency
── more on @tlc 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/no-scrubs-ai-video-f…] indexed:0 read:4min 2026-11-18 · —