• Home
  • News
  • Business
  • Why Labfresh ditched AI campaigns after four seasons: ‘Humans perform better’

Why Labfresh ditched AI campaigns after four seasons: ‘Humans perform better’

AI still evokes different reactions. Enthusiasm and trepidation has now given way to experimentation and implementation. One of the fashion companies that started working with AI early on to help with its campaigns was Dutch brand Labfresh. After four seasons with AI by its side, however, the brand is now returning to fully “human” campaigns. Co-founder Kasper Brandi Petersen gives insight into the decision.

Labfresh jumped on the AI-train quickly. What made you excited to implement it further into the operations?

At the time, I had just come home from an eight day shoot in Curacao. Before that, we tried shooting the summer collection in Denmark and Amsterdam, but it's a mess because we shoot it in January (like most fashion brands).

My friend, author and Labfresh brand board member just started writing one of the first books about AI art and had made amazing creations in Midjourney. I asked him to help us replace the big summer shoot with a Midjourney campaign and a photo shoot of real models and real products in Copenhagen. This (picture below) was the result.

The first AI campaign Labfresh ever did. Credits: Labfresh

What has Labfresh used AI for and for how long?

The above was an amazing project (and we deliberately made it identifiable as AI, so nobody felt like we tried to trick them). Unfortunately the campaign underperformed compared to other seasonal campaigns. One of the reasons was that the videos we got out of the shoots didn't work well enough (that part is much better nowadays, but hey, that's why being an early mover is better than being a first mover or pioneer). In the end it cost 12,000 euros less than a seasonal shoot, but the opportunity cost of being unable to deploy the same level of budget in print and online advertising, as we normally do, meant we pivoted.

The next season (autumn-winter) we shifted to Bloom Studios (Amsterdam), so we could skip the real photoshoot (and the annoying graphic work associated with merging real humans into AI creations).

One of the campaign images Labfresh made with BloomStudios. Credits: Labfresh

It did not work in the end. The main issue being that AI is great at creating lifestyle campaigns, but it's almost impossible to create 10 images of the same man. There's always slight differences, so it looks like 10 brothers instead of 10 images of the same man.

So for e-commerce photography we shifted a third time, to Fashionlab (Copenhagen). It's better, but e-commerce photography is actually much harder than campaign photography for AI, because the products have to be 100 percent correct in terms of fit, features, and colours. Which required a lot of work. This is why Fashionlab and most other AI software have pivoted towards requiring brands to buy bigger packages of, for example, 500 images for 7,000 euros, so the cost per image for them goes down.

What I loved about AI for e-commerce content is that we could do it on a running basis, instead of the old school way of waiting until we have enough new styles to fill up a full day of shooting models at iHeart studios.

When we have to bundle styles for AI production too, then it loses its advantages, because there is also no money to save, not for Labfresh nor for bigger brands like Mango or Zara (yet).

Labfresh has now returned to doing fully human photography. Can you explain why?

Performance: It's the same price as AI, but human results historically perform better on social media. We know this better than normal brands as we might "only" spend 1.2 million euro on Instagram this year, but we can track everything in more detail than normal brands because we are 100 percent DTC and hence have the email address of every person who ever bought Labfresh.

Price: AI photography from all the providers we sourced end up being between 15 to 100 euros per image or "moving still" video. [Editor's note: This is a still image that is afterwards animated into a short video]. We pay less than that when shooting ourselves. We get down to 10 to 15 euros per image and 50 euros per animated still.

Example: We shot AW25 this month over two shooting days. Day one, we rented a sailing boat, a bar and a fish shop. Day two, we shot on a 1,200 euro house boat for a full day. There were three creators shooting non-stop both days using big cameras, but also DJI sticks and an old school cam corder. We shot models, but also influencers and team members. We produced a crazy amount of content, that we afterwards turned into approximately 200 videos in four languages, via all sorts of AI apps.

Kasper Brandi Petersen (right) and his co-founder and partner Lotte Vink Credits: Labfresh

What would make you switch back to AI?

We won’t replace human photographers and human models in the coming two seasons. There is just not enough upside to justify further experimentation: We already took it further than any other fashion brand I have spoken to.

However, I cannot imagine my company surviving if we suddenly stopped using AI in content production. We launch more than 100 videos per month and I only have three full-timers in my marketing team. We absolutely need AI to turn fancy photoshoots into profitable campaigns because we live in a hyper-fragmented world where creative diversity, according to Meta (and myself), are the only way to make profitable campaigns on social media. This means turning images into videos, making me and my co-founder speak fluent German, adding 10 new hooks to a video, using 4,000 customer reviews to make TikTok-friendly scripts, etc.

What do you think other companies need to know about this use of AI?

AI enables you to tell better stories, but if you want to use it as a cost saving tool now is not the time. The competition for eyeballs is so fierce these days, so if you can make 50 percent cheaper content with AI, but that content converts 10 percent worse, then for Labfresh and most other brands this is not worth it. Worse engagement doesn't mean a campaign just becomes 10 percent less profitable or scalable. It means the campaign gets close to 0 percent of the budget, because the other assets in your Meta ASC campaigns will steal 100 percent of the budget as soon as the algorithm learns that the other assets are 10 percent more engaging.

This means that AI can become a megaphone for brands with engaging stories and amazing real humans representing them. Whereas boring products with no authenticity will drown in AI slop.


OR CONTINUE WITH
AI
Labfresh
Marketing