EP 073 · 78 MIN HER EXPLODING IKEA BOX SOLD BRANDS ON AI
She Cracked AI Advertising Before Brands Were Ready
The short version
Creative director Salma Aboukar tells Fast Hours how she moved from CGI in Blender and 3ds Max to an AI-native ad stack, and how her viral exploding-IKEA-box video, built on Veo 3 and a JSON prompt, convinced brands like Coca-Cola and Panasonic that AI video was ready. She breaks down her current pipeline, including Claude skills, Midjourney, Nano Banana, Z Image Turbo and Qwen LoRAs, and argues original taste, not technical skill, is the real creative moat.
Key takeaways
- 01
Salma's viral exploding-IKEA-box ad ran on Veo 3 with a prompt she converted into timestamped JSON so the box, throw pillow and product beats stayed in sync.
- 02
Going viral landed her the Panasonic, Coca-Cola and IKEA-adjacent work, but the deals overlapped so badly she'd already signed elsewhere by the time IKEA called.
- 03
Her hot take: software engineers have no taste, and AI labs chasing 'manufactured taste' would do better hiring a stylist who's never touched AI.
- 04
She never upscales AI video footage; instead she generates in 4K or 1080p and color-corrects in Premiere, since upscalers make skin 'too sharp' and obviously fake.
- 05
Her team runs a Claude skill that stress-tests every new model on release, which is how they landed on Z Image Turbo as their aesthetic base.
- 06
She trains style LoRAs on Qwen only when a big brand's legal team won't allow Midjourney; otherwise Midjourney remains her stylist and Nano Banana her editor.
- 07
Fable designed a whole on-brand email template with working social links in 15 minutes, versus a day and a half on Opus, which floored both hosts.
Receipts · said on the record
“I don't think technical people have any taste whatsoever, like software engineers or like any … I'm sorry, but like they just don't have taste.”
“I think it can be original, but it's based on the idea that you just mentioned, which is if you pour your individual unique experience, context, expertise, right, view into an LLM at a deep level, it absolutely can be original.”
“it came up with this long JSON prompt. And then I threw it in there and honestly came up with such an amazing video.”
“there's just a there's just a sheer volume of AI content. And honestly, like it's really hard to come up with like new ideas and stress test things that hasn't already been done before.”
“That one annoyed the hell out of me too. Like it's like I started like scrutinizing things that I shouldn't be scrutinizing. I was looking at like their team photo and I'm like, who took this?”
Chapters
- 0:00:00 Meet Salma Aboukar
- 0:02:17 From CGI agency to AI-first studio
- 0:04:44 Product fidelity before AI got good
- 0:07:34 The duct-taped road to photorealism
- 0:11:10 Art direction beyond basic prompting
- 0:14:28 Salma’s current AI creative stack
- 0:16:08 How she stress-tests every new model
- 0:20:20 Why color correction still matters
- 0:22:25 The IKEA video that changed everything
- 0:27:13 Going viral and handling AI backlash
- 0:30:38 Originality as the next creative moat
- 0:33:05 Can AI actually be original?
- 0:38:52 Inside an AI-native creative agency
- 0:39:55 Z-Image Turbo and aesthetic base models
- 0:43:06 From client brief to automated pipeline
- 0:47:54 Style LoRAs for brand consistency
- 0:49:11 Claude, MCP, FAL, and leaving ComfyUI
- 0:51:07 The model that cut a day to 15 minutes
- 0:55:13 Why AI content stopped feeling special
- 1:00:57 The value trapped in AI archives
- 1:02:06 Create for yourself or the audience?
- 1:04:31 Can engineers manufacture taste?
- 1:08:38 Finding inspiration outside the feed
- 1:11:57 Jackie Chan and thumbnail fuel
- 1:15:29 Final lessons from a creative trailblazer
Questions this episode answers
Who is Salma Aboukar?
Salma Aboukar is a creative director and founder (of Crates) who moved her studio from CGI in Blender and 3ds Max to fully AI-native ad production, working with brands including Coca-Cola, Panasonic and Google Labs.
How did Salma Aboukar make her viral exploding IKEA box video?
She prompted Veo 3, then had an LLM convert her detailed description into a timestamped JSON prompt so the box could explode into furniture pieces in the right order and stay seamless. The video went viral and, in her telling, was one of the moments that convinced brands AI video was commercially ready.
What AI tools are in Salma Aboukar's creative stack?
She uses Midjourney as a creative stylist, Nano Banana for editing, Qwen-trained style LoRAs for brand-specific looks, Z Image Turbo as her go-to open-source aesthetic base model, and Seedance for video, with Claude skills and FAL routed through MCP running much of the pipeline.
Does Fast Hours think AI can be original?
Drew argues AI alone is only derivative, but can produce original work when a person pours their own unique experience and taste into it; Salma agrees, adding her own hot take that most software engineers lack the taste to do this well.
Why does Salma Aboukar avoid AI video upscalers?
She says upscaling makes footage look too sharp and artificial, especially on skin, so instead she generates natively in 4K or 1080p and does color correction in Premiere Pro to keep results photorealistic.
Show notes
In Episode 73, Salma Aboukar joins the show. A creative director and founder, she explains how she moved from painstaking CGI workflows in Blender and 3Ds Max to AI-native campaigns for brands including Coca-Cola, Panasonic, and Google Labs.
She breaks down her viral IKEA exploding-room video that helped brands see the commercial potential of generative AI video, the detailed JSON prompting method behind it, and the modern AI creative stack she uses across Claude, Midjourney, Nano Banana, Seedance, FAL.ai, Z-Image Turbo, Qwen, style LoRAs, and custom AI agents.
The conversation goes deep on AI advertising, product fidelity, photorealistic skin, color correction, video upscaling, automated client workflows, and why high-end AI work still depends on original concepts, trained taste, and obsessive finishing.
They debate whether AI can truly be original, why technical teams struggle to manufacture taste, how creators survive a feed flooded with AI content, and why the next creative moat may come from the experiences, references, and strange little details nobody else can copy.
