While Baking a Cake, the Launch Site Was Finished. Just One Word: "Nice"
While baking a cake, the launch website for a new menswear collection was completed. In a marketing video for 'dots,' OpenAI's always-on AI agent, a brand operator handed the site-building task to the agent, stepped away, and on returning glanced at the result for a few seconds before saying "nice" and instructing only a few photos to be swapped. He then left again, leaving even the publishing to the agent.
A similar scene repeated in the launch video for the frontier model GPT-6 Astra. While playing tennis, a woman gave ChatGPT voice instructions to produce a presentation deck for next season's rainwear, leaving only the vague guideline "high-end and colorful," then accepted the result as is and even asked it to book a tennis lesson. The Interline flagged as a problem that these demos, which OpenAI released in two consecutive months, repeatedly show high-stakes decisions—such as product launches and buyer presentations—being handed to AI while operators react with indifference.
Citing data from a16z and its own survey, The Interline pointed out that whether AI adoption actually translates into product performance matters, and that this could become even more important as agentic commerce reduces website visits. Around the same time, OpenAI also launched a new virtual try-on service linked to its own in-app shopping experience, continuing its strategy of establishing itself as users' default operating environment.
Source · The Interline, "AI Labs Keep Demoing The Weirdest Fashion Use Cases"
$400 Million Raised, $5.5 Billion Valuation: The Report Card on Higgsfield's AI Video
AI video and image platform Higgsfield raised about $400 million (roughly 560 billion won) in funding, earning a valuation of approximately $5.5 billion (roughly 7.7 trillion won). The figure is based on annualizing revenue from its best-performing month, but the claim also surfaced that annual revenue reaches $700 million (roughly 980 billion won).
In The Interline's 2026 fashion AI survey, visual content generation (images and video) was named both the most mature area of AI use and the biggest growth area. While the performance of image generation models has recently plateaued, video generation technology continues to advance, and AI-generated content used on brand and retailer sites is already far more widespread than consumers realize. This reality is expected to come to light once upcoming mandatory AI content labeling rules take effect.
Higgsfield also applied its follow-up trained model 'Soul' to filmmaking, releasing 'Obsession,' and is expanding into fashion brands' own content production and the UGC (user-generated content) market. The Interline also identified adding AI vision and analytics technology to existing camera footage from stores and manufacturing floors as another growth axis, but added that errors—whether in generation or analysis—could drive up costs.
Source · The Interline, "Two Takes On How AI And Video Are Intersecting In Fashion"
Video Footage Experiments Starting From a Single Image: RealDrop Clip
RealDrop's Clip is a feature that creates a short video by adding effects and a prompt to a single image. When you have lookbook or product shots but no video footage to post on Reels or TikTok, you can use the same image to create several differently directed clip versions side by side and compare them.
Writing the desired motion or mood as a sentence in the prompt field lets you try clips with different feels even from the same image.
Lining up and playing back the multiple clips filling one row lets you compare versions to use as ad creative.
One approach is to use RealDrop's Clip to expand a single lookbook image you already have into multiple clip versions, then move the cut you like best into a Reels or ad asset.
