The brick by brick AI trend is a 6-second video in which you replace Franklin Saint, the driver in Snowfall's "I built this, me, brick by brick" car speech, with yourself. On videosi you upload one photo and get the clip with the original dialogue on it; there is no prompt to write.
This guide goes deeper than the FAQ on the trend page. It covers the original scene, why the line spread, why close-ups are hard for AI, and what to do about subtitles, labels and cost. It is an early trend with little AI-specific evidence, and videosi is not affiliated with the show or its makers.
Key takeaways
- The clip is 6 seconds long and carries the original dialogue, ending right after "tear it down"; your lips will not match the words exactly.
- The reference clip has English subtitles burned in. videosi has them removed, but a trace can occasionally remain.
- The line "I built this, me, brick by brick" has been a sound since at least June 2023; the AI swap versions only started around early November 2026.
- A tight close-up makes likeness hard, so use a sharp, front-facing photo.
- It costs 10 credits per video on videosi, with packs from $8.90 for 10 credits.
- This is an early trend, and videosi is not affiliated with the show or its makers.
In this guide
- What actually happens in the original car scene?
- Why did one line from that scene become a meme?
- How does an AI character swap work in this scene?
- Why is a close-up so hard for AI to get right?
- Why do subtitle remnants show up, and how are they removed?
- Which photo gives the best likeness?
- Why is the clip only 6 seconds, and where does it stop?
- Does the video have sound, and do the lips match?
- How do I label an AI video, and why does it matter?
- How is this different from a CapCut green-screen template?
- How many steps does it take with a generic video tool?
- How big is this trend, and how sure are we?
What actually happens in the original car scene?
The scene is a conversation inside a car in the crime drama Snowfall. A young man at the wheel turns on the passenger next to him and gives a short, fierce speech about something he made himself. The speaker is Franklin Saint, played by Damson Idris.
What makes it work as a template is how little there is. One location, one face, one hand jab, one speech. The camera looks at the driver almost the whole time, so everything the viewer feels comes from the head turn, the stare and the proud, angry smile.
Why did one line from that scene become a meme?
The line is short, rhythmic and easy to borrow: "I built this, me, brick by brick". It works over any boast about something you made slowly and yourself, such as a career, a channel, a body or a business. That makes it flexible, which is what memes need.
The sound has circulated since at least June 2023, mostly in dark, slow edits that quote the whole speech, and several of those edits passed 3 million views. A green-screen version also circulates as a CapCut meme template. So by the time AI versions appeared, people already knew the sound by heart.
How does an AI character swap work in this scene?
A character swap takes a real video and replaces one person in it with someone from a photo. The original clip supplies the motion: the camera, the turn of the head, the hand jab and the pacing. Your photo supplies the face, hair, build and outfit.
That is how the brick by brick AI page works. You give one photo, the AI puts you in the driver's seat, and the passenger, the car and the daylight stay as they were. Nothing is filmed. The whole result is generated and re-edited, which is why it should be labeled as AI.
Why is a close-up so hard for AI to get right?
In a wide shot the face is small, so small errors hide. Here the face fills the frame, and the viewer sees every tooth, the jaw line and each change in expression. Any drift in likeness is obvious.
There is also a pull toward the original actor. When the replacement is not forced hard enough, the model quietly keeps the original face, and you get a result that looks like someone else. Lip movement and the head turn can also drift when a face is replaced without a real motion reference. The speech has fast, forceful mouth shapes, so a loose reference shows quickly.
Why do subtitle remnants show up, and how are they removed?
Popular clips of this scene usually have captions burned into the picture. A model that treats the video as a reference may copy them into the result as garbled, shifting text, which looks like a glitch rather than a caption.
The fix is to tell the model to remove all subtitle text while keeping everything else in the scene the same. The page does this for you, but now and then a render can still keep part of a word or a few garbled letters. Watch the whole clip before you post, and generate again if you see text (a new render uses credits again). If you work with a generic tool, put the same instruction in your prompt and check every try for leftover letters.
Which photo gives the best likeness?
Use a sharp, front-facing head-and-shoulders photo with your whole face visible. The scene is nearly all face, so the photo decides the result. A distant full-body shot leaves the model to invent the face.
A photo where you look at the camera or just beside it matches the stare in the first half second. Avoid sunglasses, masks, hats and heavy filters, since a clear forehead and eyes keep the face stable while the head turns. A plain dark top looks natural in the warm, dim light of the scene. If the clip still shows the original actor, regenerate once; a failed render returns its credits automatically.
Why is the clip only 6 seconds, and where does it stop?
The videosi clip is 6 seconds long and ends right after the line "I'll be damned if I let you tear it down". The scene in the series keeps going past that point, but the clip stops there. That also keeps the clip short enough to post as a Reel, a Short or a TikTok without trimming.
Inside those 6 seconds there are five beats: the stare, "I built this", the sharp nod on "me", the lean-in on "brick by brick", and the voice rising into an angry laugh. Knowing the beats helps you pick a photo whose expression fits the end of the clip.
Does the video have sound, and do the lips match?
Yes, the video has sound. Every brick by brick render comes with the original dialogue from the series, the actor's own voice, so you can post it as it is.
The lips do not match exactly. Your face is generated to follow the head turn and the mouth shapes of the scene, but the voice is not yours, so treat the result as a dub rather than a lip-synced performance. If you prefer, you can swap the sound in the app you post to.
How do I label an AI video, and why does it matter?
Many platforms ask you to mark realistic generated video as AI. Check the settings of the app you post to, and add the tag or a plain line in the caption. Without it, people may take the clip for the real show.
The sound is the original dialogue from the series, so a platform may mute or flag it on some accounts. Nobody can promise how a given post will be handled. Use photos of people who agreed, and never use the clip to make it look like a real person said something they did not.
How is this different from a CapCut green-screen template?
The popular CapCut brick by brick templates are green-screen meme templates. They drop your footage or photo into an existing layout. They are not an AI swap, and they change or disappear often.
The AI version regenerates the scene around you, so the head turn, the lighting and the speech all fit your face. There is no live camera filter for it. If you want the AI version, upload one photo to the brick by brick AI generator and download a 6-second MP4.
How many steps does it take with a generic video tool?
You can do it yourself, but it takes more work. You need the source scene as a motion reference, a photo, a prompt that replaces only the driver, a rule to remove subtitles, and a way to put the original audio back on the result.
Then come the retries. Each try costs credits and several minutes, and most people need a few before the face holds and the timing matches. On videosi the same job is one photo and one button: 10 credits for a 6-second clip, credit packs from $8.90 for 10 credits and plans from $29.90 a month.
How big is this trend, and how sure are we?
Not big yet. The sound is old, but the AI character swap is new: the first posts we found appeared around 5 to 7 November 2026. We found only a few, with small view counts, and we could not confirm who made the first one.
So treat it as an early trend with little AI-specific evidence. It may grow or fade quickly. videosi is independent and has no connection to the series or its makers, and the page only states what we could verify.



