Anyone Can Make Insane Visual Effects Now!
Matt Wolfe opens the full toolbox behind his own YouTube intros — every AI app, model, and exact prompt, from wall-bursting entrances to a fully AI-dramatized Sam Altman text leak.
July 8thA motion design educator runs Google's upgraded Omni/Flow model through a real ad brief, then argues his own code-based Remotion pipeline still wins on control.
Google's upgraded Omni model turns one detailed prompt into a branded 10-second ad, but real creative control still belongs to code-based generation tools like Remotion.
Fernando Araujo tests Google's upgraded Omni video model by building a mock ad for a fictional travel-card brand and comparing it to his own code-based Remotion workflow. A vague one-line prompt produces generic, forgettable footage, but feeding the generator a fully detailed five-scene shot list, written by his own AI 'director' agent, produces a coherent, on-brand result. He refines it further: uploading his own photo as a reusable character, swapping in the real product design, and adjusting an on-screen mockup through follow-up instructions. His conclusion: text-to-video tools win on realism and speed but only let you steer results through more instructions, while text-to-code tools trade some realism for exact, repeatable control.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →
States the premise: Google updated its video model, so he's testing what it can do for design work, not just realism. Shows an earlier Coca-Cola-style test clip and mentions a prior Nubank test where the model missed the brand's visual identity.

Walks through the refreshed Flow interface (new project screen, 'Motion 4x faster' messaging), then uploads a personal photo to create a persistent AI character named 'Brabo' with a chosen voice and personality.

Gives the model a short, generic brief for a 5-scene ad for the fictional 'Nomad' travel card. The model returns a montage of stock-photo-style people in vague travel/finance settings. Notes the 10-second, 4-variation generation costs about 120 credits.

Re-runs the prompt with his uploaded character tagged as the protagonist, describing a mini airport/payment story. Result is more specific but still thin, so he introduces his own AI 'director' agent (built for his Remotion-based course) that writes fuller creative briefs.

Prompts Diretor Brabo for a complete Motion Design brief in the brand's yellow/black/white palette: camera direction, scene-by-scene descriptions, on-screen interface elements, and card presentation. Feeds the resulting script back into Omni.

The scripted brief produces a coherent branded spot ('Sua vida global começa aqui'). He then refines it with follow-up instructions: swapping the black card render for the real yellow card image, adding a 3D reflection pass, and replacing the phone-screen mockup content.

Draws the video's central distinction — text-to-video generates pixels you steer with instructions, text-to-code generates a program you can edit directly — then finalizes the logo reveal animation and closes with a nod to his AI motion design course.
The gap between a forgettable AI ad and a convincing one is almost entirely the amount of visual detail in the prompt, plus knowing whether the job calls for pixel generation or code generation.
“Cara, o resultado disso aqui está muito incrível, galera.”
“já deixei aqui, vai gastar 120 créditos para gerar 1 vídeo de 10 segundos e gerar 4 vídeos é, eu tenho 1 plano ultra.”
“de vídeo, ele está gerando vídeo, ele não está escrevendo 1 vídeo. O ele está escrevendo 1 vídeo, por isso que a gente tem o código, a gente consegue ir lá no código mudar alguma coisa.”
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
Google quietly upgraded its video-generation model, and this motion designer's first move was to stress-test it against real ad work rather than a demo reel. He starts with a one-line prompt for a fictional travel card, watches it come back generic, then rebuilds the same idea through his own AI 'director' agent, a reusable character, and a string of targeted edits until it looks like something a real brand could ship.
The video's central distinction between two AI video paradigms — generating pixels directly (Omni) versus generating an editable program that renders pixels (Remotion via his 'Diretor Brabo' agent).
“esse esse agente aqui diretor, ele faz parte do do curso comerciais com IA, onde eu ensino as pessoas a criar com inteligência artificial só que usando o Remotion”
Soft, in-context plug woven directly into the workflow explanation rather than a hard sales break — the course comes up exactly when he introduces the tool it produced.
00:01
00:25
00:42
00:59
01:16
01:33
01:49
02:03
02:23
02:40
02:57
03:14
03:31
03:48
04:05
04:22
04:39
04:55
05:12
05:29
05:46
06:03
06:20
06:37
06:54
07:11
07:28
07:45
08:02
08:18
08:34
08:52
09:09
09:26
09:43
10:00
10:17
10:34
10:51
11:08
11:24
11:41
11:58
12:15
12:32
12:49
13:06
13:23
13:40
14:02
14:07
14:30
14:50
15:04
15:13
15:38
15:55
16:05
16:29
16:46
17:03
17:20
17:44
17:49
18:10
18:21
18:44
19:01
19:18
19:35
19:52
20:10
20:26
20:43
20:59
21:16
21:33
21:50
22:07
22:16Matt Wolfe opens the full toolbox behind his own YouTube intros — every AI app, model, and exact prompt, from wall-bursting entrances to a fully AI-dramatized Sam Altman text leak.
July 8thA screen-recorded walkthrough of Open Generative AI, an open-source clone of Higgsfield that runs image and video generation on your own hardware for free instead of a paid subscription.
May 9thA screen-recorded walkthrough of building a custom Claude Code skill that watches any viral video, breaks it into timestamped beats, and recreates it with Higgsfield's Seedance 2.0.
July 6thMKBHD's full post-production team — lead editor, colorist, motion designer, sound designer, and audio engineer — reveals the craft decisions behind every frame.
October 12th 2025A 19-minute deep-dive into the updated AI cinema pipeline — three changes that killed the plastic look for good.
May 27thA 16-minute walkthrough of an AI affiliate pipeline that turns a single Amazon product image into a postable UGC short — no camera, no editing skills, no scripting from scratch.
June 15th