Latest / The DTC Marketing Podcast with Fexingo: Direct-to-Consumer Brands, Shopify, and E-Commerce Growth / How DTC Brands Use AI Voice Cloning for Personalized Marketing
Transcript
- Lucas: So you're a DTC brand, you've got a founder with a distinctive voice — maybe it's warm, maybe it's quirky — and you know that when your customers hear that voice in a video or a voicemail, they're more likely to buy. The problem is, your founder can't record a thousand personalized messages a day. Luna: Right, and that's where AI voice cloning comes in. We're not talking about deepfakes for scams — we're talking about legitimate marketing use cases. Lucas: Exactly. There's a company called ElevenLabs that's been getting a lot of attention in the DTC space. They let you clone a voice from a short sample — like five minutes of audio — and then generate new speech from text. For a brand, that means you can have your founder "say" a different message to every single customer. Luna: I've seen a mattress brand do this. They cloned the founder's voice, ran retargeting ads where the ad copy changed based on what the user had browsed, and the click-through rate jumped 40 percent. Lucas: Forty percent — that's huge. And the cost difference is stark. A single AI voice clone subscription runs about ninety-nine dollars per month. Compare that to a studio session for a real voiceover — you're looking at five hundred to two thousand dollars per hour, plus the founder's time. So you can iterate fast. Luna: But there's a trust question here. If I get a voicemail that sounds exactly like the founder, but it's actually an AI — do I feel manipulated? Or do I feel like the brand cares enough to personalize? Lucas: That's the line. And smart brands are handling it by being transparent. Some include a brief disclosure at the start — something like, 'Hey, this is an ai generated message from our founder.' They've tested it and found that disclosure doesn't hurt conversion, but it protects against backlash. Luna: I think it's smart. You're essentially trading a small dip in surprise for a big gain in long-term trust. And the data backs that up — consumers are increasingly wary of undisclosed AI. Lucas: Yeah, and let's be honest — a lot of what makes DTC brands work is the personal touch. The founder story, the authenticity. If you lose that, you're just another commodity seller. So cloning the voice actually preserves that connection at scale. Luna: Quick honest thing — a handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that's literally what funds making episodes like this one. Keeps it ad-free and independent. Lucas: Yeah, it's been that way from the start. No sponsors, no annoying mid-rolls — just listener support. If these conversations have sparked something you've actually used, that's the link. Luna: Alright, back to voice cloning. So we've got the cost and trust angle — what about the actual implementation? How does a brand even set this up? Lucas: Let's walk through a real example. There's a supplement brand — call them VitaDirect — that started using ElevenLabs for their abandoned cart sequence. They had the founder record a five-minute sample, then they generated three different versions of a follow-up message: one for cart abandoners, one for people who browsed but didn't add, one for repeat buyers. Luna: And they A/B tested it against a generic text email? What happened? Lucas: The voice messages drove a 25 percent higher conversion rate on abandoned carts compared to the text-only email. And the cost per conversion dropped because there was no incremental production cost per variant. Once the clone is made, you can generate a hundred different scripts for free. Luna: That's the scalability piece. But I'm curious about the creative side — does the AI capture subtle emotion? Like, can it sound genuinely excited about a new flavor? Lucas: It's gotten surprisingly good. ElevenLabs offers something called "speech to speech" where you can have the AI read a script in a specific emotional tone — excited, calm, urgent. You can even adjust the pacing. So the founder records a few sentences in an excited voice, and the model learns that style. Then you can generate new excited messages. Luna: So it's not just a flat reading — it's nuanced. That's where I think it becomes a real marketing tool rather than a gimmick. Lucas: Right. Another example: a pet food company started using voice clones to create personalized video thank-yous after purchase. They'd include the customer's pet's name and a specific mention of the product bought. The videos were shared on social media organically — people love showing off that their dog got a shout-out from the brand. Luna: That's smart user-generated content that's actually brand-generated. And it's cheap. What about the risk of misuse? Say a disgruntled employee gets access to the clone and generates something harmful. Lucas: That's a real concern. ElevenLabs and other providers have been adding safeguards — you can set permissions, limit the number of generations, require approval for each script. But ultimately, the brand is responsible. If you're cloning your founder's voice, treat that asset like you'd treat your brand logo — control who can use it. Luna: And there's the regulatory side. The FTC has already signaled they're looking at ai generated content in advertising. If a consumer can't distinguish a real founder from a clone, that could fall under deceptive practices. Lucas: Yeah, I think we'll see a standard emerge: any ai generated voice in a commercial context must be labeled. Some brands are already doing it voluntarily. And honestly, it might become a badge of trust — 'Hey, we're using AI, but we're upfront about it.' Luna: Let's talk about the economics a bit more. You mentioned ninety-nine dollars a month for a clone. What about the text to speech generation costs? Lucas: ElevenLabs charges based on characters processed. For a typical DTC brand doing a few thousand voice messages a month, you're looking at maybe an extra fifty to a hundred dollars. So total voice costs under two hundred dollars a month. Compare that to hiring a voice actor for a single campaign — you'd pay that for one hour of studio time. Luna: So the ROI is almost immediate if you're doing any volume of personalized outreach. And the quality is good enough that listeners don't immediately clock it as AI. Lucas: Exactly. One thing I've noticed is that the best use cases are where the voice is expected — like a voicemail or a video message. If you suddenly have a robot voice in an ad that's always been human, that's jarring. But if you're building a new channel, you can start with AI from day one. Luna: What about languages? Can you clone a voice in multiple languages? Lucas: Yes — ElevenLabs supports something like twenty-nine languages. You can have the founder's voice speaking Spanish, French, Mandarin, all from the same clone. That's huge for DTC brands expanding internationally without hiring local voice talent. Luna: But the clone might have an accent. Is that a problem? Lucas: It depends. For some brands, the founder's accent is part of the charm — think of a Scottish founder selling whiskey. For others, you might want a neutral accent. You can actually train the clone to sound more native by providing samples in that language. It's not perfect yet, but it's improving fast. Luna: So where do you see this going in the next twelve months? By June 2027? Lucas: I think we'll see real-time personalization. Imagine a customer visits your site, and the chatbot greets them with the founder's voice, using their name and referencing their last purchase. That's already possible technically, but brands are still testing the waters. The ones that move early will own the customer relationship. Luna: And the ones that don't? They'll be competing on price and features — which is exactly the race to the bottom DTC brands try to avoid. Lucas: Right. Voice cloning is a way to differentiate on experience, not on product specs. And in a crowded market, that's everything.