TTS

❶ Can AI voices really sound natural enough that customers won’t realize they’re talking to a machine?

This is the first question most restaurant operators ask, and the short answer is: today’s technology can get remarkably close.

The key difference comes down to which generation of TTS technology you’re using. Older concatenative TTS did have that unmistakable robotic quality. But modern neural TTS works differently. It’s built on deep learning models that capture the subtle details of human speech—the natural rise and fall at the end of a sentence, the slight emphasis on key words, even well-placed pauses.

What we’ve observed with Tunvo deployments is an interesting pattern: when the voice sounds natural enough, customers simply don’t stop to think about whether they’re talking to a human or an AI. They just place their order. It’s only when the voice has obvious mechanical artifacts that customers become guarded.

That said, expectations should be realistic. In a quiet office, with high-quality audio, some people might still spot the difference. But in real restaurant environments—with background noise, customers calling from cars or busy streets, and the natural compression of phone lines—a well-tuned TTS voice passes as human for the vast majority of callers.

What matters for restaurants: Don’t just evaluate voices through laptop speakers. Test them on actual phone lines. Tunvo works with clients to conduct proper channel testing, ensuring the voice that sounded warm in a demo maintains that quality after it’s passed through a real telephone network.

❷ How does TTS handle unusual menu items or branded dish names? What if it mispronounces something?

This is a practical concern that directly affects brand professionalism.

Imagine a customer calls to order and the AI reads “Chicken Parmesan” as “Chicken Par-mee-see-an” or stumbles over a proprietary pizza name. Customers may not complain, but it creates a subtle impression that something is off.

The industry term for solving this is deterministic pronunciation control. Tunvo’s approach is straightforward: instead of letting the AI guess how to pronounce things, restaurants teach the system the correct way.

Here’s how it works in practice:

• Before launch, Tunvo works with the restaurant to review the menu and flag any items that need special handling

• The restaurant provides the correct pronunciation—this can be a recording, or using phonetic notation

• These pronunciation rules go into a dedicated lexicon, and the system follows them every time

This has a practical advantage for menu updates. When new items roll out, they can be added to the lexicon immediately. No waiting for model retraining. Today’s new special, pronounced correctly today.

A restaurant operator once put it this way: “With generic TTS, we used to cross our fingers and hope it didn’t mangle our signature dish. Now we don’t have to hope. We know it’s right.”

❸ Can we customize the voice to match our restaurant’s brand personality?

Absolutely—and this is where modern TTS really shines.

TTS isn’t a one-voice-fits-all technology. Quality systems offer multiple voice options, and within those options, fine control over speaking rate, pitch, and tone.

Tunvo helps restaurants develop what we call a voice persona:

• Family restaurants tend to work well with warm, approachable voices—the kind that feels familiar, like a server who’s been there for years

• Quick-service brands often prefer crisp, efficient voices that match the speed of the experience

• Fine dining establishments typically go for voices with more gravitas, a measured pace that suits the setting

• Youth-oriented brands might choose energetic voices, sometimes with a touch of playfulness

Beyond the base voice, there’s scenario-based tuning. When confirming a reservation, the tone can be slightly upbeat. When delivering bad news—no tables available—the voice can soften, slow down, convey genuine regret. These small adjustments make the AI feel more aware of context.

There’s a saying at Tunvo: get the voice right, and the technology fades into the background. Get it wrong, and no amount of technical capability will fix it. Because for many customers, that voice is the first real contact they have with your brand.

❹ Won’t customers find an AI voice cold and impersonal? Hospitality is about human connection.

This concern comes from a good place—and it deserves a thoughtful answer.

But the reality might surprise you. The issue isn’t really whether the voice comes from AI or a human. It’s whether the interaction works.

What Tunvo has observed: when the voice sounds natural, the script is well-written, and the flow makes sense, customers have positive experiences. They get their orders placed. They get the information they need. They accomplish what they called to do. That is service.

The alternative isn’t a warm human every time. The alternative is often a busy signal, or a rushed employee juggling three things at once, or a frustrated customer put on hold. That’s what actually feels cold.

Consider the Donatos Pizza case. After deploying AI voice ordering, customers got through faster and placed orders more smoothly. Satisfaction didn’t drop—it improved. Not because AI is better than a great human server, but because AI never has an off day. Every call gets the same consistent, patient service.

That said, Tunvo encourages restaurants to invest in script design. Small touches matter—a well-placed “let me check on that for you,” a genuine-sounding “I’m sorry, we’re fully booked right now.” And when the AI can’t solve a problem, a smooth handoff to a human who already knows what’s been discussed makes all the difference.

The short version: Good TTS doesn’t replace warmth. It delivers warmth consistently, at scale, without fatigue.

❺ How will our staff react to AI taking phone calls? Won’t they feel threatened?

This is as much about change management as it is about technology. Get it right, and everyone wins.

Look at the numbers from Donatos: after launching AI voice across 174 locations, they reclaimed nearly 5,000 labor hours in a single month. Those hours didn’t disappear—they got redirected. Staff spent less time tethered to phones and more time on food prep, table service, and taking care of in-person guests. Many employees actually found the work more satisfying.

From Tunvo’s experience working with restaurant chains, the most successful implementations share a few practices:

• Communicate early and clearly: Before launch, explain to the team what the AI will do and—just as important—what it won’t do. It’s there to handle the repetitive, high-volume stuff, not to replace people.

• Define clear boundaries: AI handles routine orders, reservations, and simple questions. Complex situations, emotional customers, anything needing on-the-ground coordination—those go to humans.

• Frame it as opportunity: Some restaurants show their teams the math: “This many hours saved from phone duty means we can serve this many more tables during rush.” It reframes the conversation from loss to gain.

One store manager put it well: “Before, during the lunch rush, I was getting pulled in three directions—phones ringing, kitchen calling for orders, line out the door. Now the phones are covered. I can actually focus on the dining room. It’s way less stressful.”

The point isn’t AI versus humans. It’s AI handling what it’s good at—repetition, scale, consistency—so humans can do what they’re good at: connecting with regulars, handling the unexpected, bringing warmth to the moments that need it.