Recreating Simon Meyer's Miniature VW Bus Commercial | With Commentary | GPT-5 & Veo 3

Recreating Simon Meyer's Miniature VW Bus Commercial | With Commentary | GPT-5 & Veo 3

In this video, I test if modern AI tools can recreate Director Simon Meyer's viral miniature VW Bus commercial. I use only GPT-5, Topaz, and Veo 3 to recreate the commercial's distinctive tilt-shift photography effect and forced perspective, then analyze the results.

What are the pluses and minus of using generative AI versus Unreal Engine to do this?

COMMENTARY: Do you remember when making a commercial like this cost hundreds of thousands of dollars? Simon Meyer created an ultra-realistic miniature car commercial and shared the prompts on Instagram. This is my re-creation using only GPT-5, Topaz, and Veo 3.

GPT-5 is slow, but it remembers that all of your images are related, which keeps consistency across shots — for example, maintaining the correct proportion between the hand size and the VW Bus. Midjourney fails to do that. However, one can use GPT-5 images as references for Midjourney, and the consistency improves. I will start using GPT-5 for storyboarding, even if those images are only used as references.

Veo 3 limited me to five videos despite my very expensive Ultra subscription. Google does not design their tools for real video production; they treat Veo 3 like a toy — even watermarking every image whether you want it or not. Its strength is adding sound, but Kling is probably better at sound. Veo 3 is better at the physics. The other tools moved the VW bus in unnatural ways.

🔍 EXACT PROMPTS USED: I used are at this site https://www.nikbearbrown.com/vw-bus

African Savanna: "An oversized hand moves a 1960 Volkswagen Kombi Bus in white and pink with silver fenders and German license plates through an African savanna with a startled giraffe. Tilt-shift photography technique creates a distinctive depth of field that makes the scene appear miniaturized and toy-like..."

Stormy Coastal Cliffside: "An oversized hand moves a 1960 Volkswagen Kombi Bus in white and pink with silver fenders and German license plates along a stormy coastal cliffside. Tilt-shift photography technique..."

Underwater Coral Canyon: "An oversized hand moves a 1960 Volkswagen Kombi Bus in white and pink with silver fenders and German license plates through an underwater coral canyon..."

Snow-Covered Alpine Pass: "An oversized hand moves a 1960 Volkswagen Kombi Bus in white and pink with silver fenders and German license plates through a snow-covered alpine pass..."

Nighttime Cityscape: "An oversized hand moves a 1960 Volkswagen Kombi Bus in white and pink with silver fenders and German license plates through a glowing nighttime cityscape..."

Indoor Shelf Scene: "An oversized hand moves a 1960 Volkswagen Kombi Bus in white and pink with silver fenders and German license plates on a wall-mounted shelf in a personal living space..."

Weimaraner Dog at Burning Man: "A sleek Weimaraner dog, seen from behind, runs toward a 1960 Volkswagen Kombi Bus in white and pink with silver fenders..."

FINDINGS: GPT-5 maintains consistency across shots but requires Topaz for upscaling. Veo 3 excels with sound and physics but has limitations for serious production. I'll be rebuilding this spot with GPT-5, Midjourney, Higgsfield, Veo 3, and Kling, comparing what each can do — their strengths, weaknesses, and the edges where they still need human hands. In conclusion, GPT-5 requires Topaz to upscale, but it's the consistency winner. Until Google takes AI film seriously and allows AI filmmakers to really use the tool, it will remain a TikTok toy.

Also, I'll take responsibility for the bad edits — that part's still on me. But even as production evolves, this means that great direction is more valuable than ever. This video is about testing what's now possible when the tools change everything.

Transcript

A lightly formatted transcript of the video, with timestamps so you can jump to any section on YouTube.

[0:00] Do you remember when making a commercial like this cost hundreds of thousands of dollars? Director Simon Meyer created an ultra realistic miniature car commercial and shared the prompts on his Instagram. This is a recreation of that video using only GPT5, Topaz, and VO3. GPT5 is slow, but it remembers that all of your images are related, which keeps consistency across shots. for example, maintaining the correct proportion between the hand size and the VW bus. Midjourney fails to do that. However, one can use GPT5 images as references for midjourney and the consistency improves. I will start using GPT5 for storyboarding even if those images are only used as references. V3

[0:51] Limited me to five videos despite my very expensive Ultra subscription. Google does not design their tools for real video production. They treat VO3 like a toy, even watermarking every image, whether you want it or not. Its strength is adding sound, but Cling is probably better at sound. VO3 is better at the physics. The other tools move the VW bus in unnatural ways. I'll be rebuilding this spot with GPT5, Midjourney, Higsfield, VO3, and Cling, comparing what each can do, their strengths, their weaknesses, and the edges where they still need human hands. In conclusion, GPT5 requires Topaz to upscale, but it's the consistency winner. Until Google takes AI film

[1:42] Seriously and allow AI filmmakers to really use the tool, it will remain a Tik Tok toy. Also, I'll take responsibility for the bad edits. That part's still on me. But even as production evolves, this means that Great Direction is more valuable than ever. This video is about testing what's now possible when the tools change everything.