Latest / The DTC Marketing Podcast with Fexingo: Direct-to-Consumer Brands, Shopify, and E-Commerce Growth / How DTC Brands Use AI Product Photography to Cut Costs
Transcript
- Lucas: So there's this moment in every DTC founder's life where they're sitting on a new product, they know it's good, but they can't launch because the photography isn't ready. And traditionally, that means booking a studio, hiring a photographer, maybe a model, and waiting two to three weeks for edited images. Luna: Right. And that's easily a couple thousand dollars for a basic shoot, and five figures if you want lifestyle shots. Lucas: Exactly. But this year, a mattress-in-a-box brand — I'm not going to name them because they're still in stealth with this — they replaced their entire lifestyle photoshoot pipeline with ai generated imagery. We're talking hero shots, bedroom scenes, even the 'unboxing moment' you see on their homepage. Luna: And you said they saved how much? Lucas: Over forty thousand dollars per campaign. That's not a small number for a brand that's maybe doing two to three million in revenue. And here's the key — their conversion rate didn't drop. In fact, it went up slightly because they could A/B test different scenes much faster. Luna: That's wild. Because I think a lot of marketers instinctively push back on AI imagery — they worry it'll look fake and hurt trust. Lucas: And that worry is legitimate. Early AI product photos had that uncanny, glossy look — everything looked like a 3D render from 2015. But the models now, especially the ones fine-tuned on e-commerce data, they're producing images that pass the blink test. You have to zoom in to tell it's not a real photograph. Luna: So how do they actually do it? Like, what's the workflow? Lucas: So the brand I mentioned uses a tool that starts with a real shot of the product on a plain background — just a clean, well-lit photo. Then they feed that into a generative model with a text prompt: 'queen size mattress in a bright Scandinavian bedroom, morning light, minimal decor.' The model composites the product into that scene, matching lighting and shadows. Luna: And it gets the scale right? Because that's always been the problem with early composites — the mattress would look like a twin in a king-size room. Lucas: That's where the fine-tuning comes in. They trained the model on their own catalog images first, so it understands the exact dimensions and proportions. Now they can generate thirty different lifestyle scenes in an afternoon, then pick the best five. Compare that to the old process: one shoot, maybe three scenes, and you're locked in for months. Luna: I want to push back a little, though. If every DTC brand starts using AI photography, won't the visual landscape become homogeneous? Everything starts looking like the same Pinterest board. Lucas: That's the risk, and I've seen it happen. Some brands get lazy — they just type 'cozy living room' and accept whatever the model spits out. But the smart ones treat it like any creative tool: they build a style guide, they curate the outputs, they add their own brand elements like color grading or specific props. Luna: So the differentiator becomes curation, not the raw generation. Lucas: Exactly. The brand I mentioned has a creative director who still art-directs the AI — she'll iterate on prompts ten or fifteen times until the scene feels right. And they still shoot their core hero images traditionally. The AI is for scale and iteration, not for replacing the hero shot entirely. Luna: That makes sense. It's more of a line extension than a full replacement. Lucas: And this is where it gets really interesting for bootstrapped founders. If you're launching a new product and you don't have five grand for a photoshoot, AI photography lets you get to market with decent imagery for maybe a hundred dollars in compute costs. Then you can reinvest the savings into actual customer acquisition to validate the product. Luna: And if the product takes off, you can always go back and do the premium shoot. Lucas: Right. A lot of founders treat photography as a fixed cost they have to absorb before they know if the product works. This flips that — you can test the visual proposition with minimal risk. Luna: Speaking of minimal risk — we've been doing these episodes for a while now, and one thing that keeps this show going is that listeners like you chip in a little to keep it ad-free. If these marketing conversations have sparked something you've actually used, a couple of dollars a month is genuinely what keeps these going. Lucas: Yeah, it's at buy me a coffee dot com slash fexingo. No pressure at all, but it makes a real difference — helps us cover hosting and research time. Luna: And we really appreciate it. Now, back to AI photography — one thing I haven't seen discussed much is how this affects model diversity and representation. Lucas: That's a big one. If you're generating lifestyle scenes with people, the model can encode biases from its training data. We've seen cases where a prompt for 'young professional in a modern apartment' overwhelmingly generates white, thin, able-bodied individuals. Luna: So brands need to be intentional about prompting for diversity. Lucas: Absolutely. Some brands are building their own fine-tuned models with diverse datasets to ensure representation. Others are consciously over-sampling underrepresented groups in their prompts. It's not a set-it-and-forget-it tool — you have to actively manage the outputs. Luna: And there's the legal side too. If the model was trained on copyrighted images, could the brand be liable? Lucas: That's the open question right now. Most commercial AI image tools claim they have indemnification for generated outputs, but the case law is still developing. Smart brands are keeping records of their prompts and the model version used, just in case. And they're avoiding generating anything that resembles a known photographer's style or a trademarked setting. Luna: So for a DTC founder listening right now, what's the one concrete step they should take this week to explore AI photography? Lucas: Take your best-selling product, shoot it on a white background with good lighting — even with your phone — and run it through one of the major AI image generators with a simple prompt. See what comes out. Try five different scenes. If even one of them looks good enough to use on a product page or a social ad, you've just saved yourself a shoot. That's the test. Luna: And if it looks terrible, you're out maybe ten bucks and an hour of time. Lucas: Exactly. The barrier to experimentation has never been lower. And the brands that learn to blend AI efficiency with human curation are going to have a real cost advantage over the next year. Luna: I think that's the key takeaway — it's not about replacing creativity, it's about removing friction so you can iterate more. Lucas: Right. And that's what we're seeing across DTC right now — the winners are the ones who use AI to accelerate the loop between idea, test, and learning.