How Do You Record IVR Prompts? What I Learned Comparing Short-Form Voiceover, Podcast Intros, and Phone Greetings

Last month my cousin called me in a mild panic. She runs a flower shop, and the greeting on her phone system — the thing every single customer hears before a human picks up — was twelve seconds of her reading off a sticky note. She'd recorded it three times. All three landed somewhere between a school open-house announcement and a car dealership ad, and she was pretty sure callers were hanging up before the menu even finished.

She was probably right. People judge a business by its phone voice faster than by its website. That's just how it works.

Here's the odd part, though. Narrating a short video and recording a podcast intro look like the same job on paper. You talk into a microphone, that's it. In practice they're completely different sports, and a phone greeting is a third sport entirely.

I learned this the slow way.

Short-form voiceover: the "reading off a page" problem

Two months ago I did the narration for a cooking channel's Shorts. The script was maybe eighty words. It took me nearly two hours.

Not because my voice was unclear. Because the words had to land on the pictures. The pause before "and now the secret ingredient" had to hit at the exact moment the butter hit the pan. Half a second late and the whole thing feels off, and viewers swipe. Fifteen seconds has to carry a hook, a payoff, and a call to action, which is already a lot, and the second you tense up it turns into a late-night infomercial.

Then there's the music. I can't count how many times I've recorded a clean take, dropped it into the edit, and watched the background track eat half my consonants. Back to the mic.

Platform moderation adds its own little dance too. Certain words can't be said, so you rewrite around them, re-record, re-edit... by take six you honestly can't tell if it's any good anymore. You've heard it too many times.

One thing that has genuinely gotten easier: when a creator wants to run the same video for a Spanish-speaking or German audience, free AI dubbing built on multilingual TTS free tools can save an entire re-shoot. Same timing, new language. Two years ago that was a luxury. Now it's a Tuesday.

Podcast intros are a different animal

A podcast intro doesn't have to sync with anything. Sounds easier, right? It's harder.

There's no music bed hiding your mistakes, no cuts to mask a flubbed line. You're asking for one continuous read with what radio people call the late-night feel — that sense of one person talking to one listener. The problem is that alone in a quiet room, most of us drift into newsreader mode. Stiff. Announcy. Like we're reading the six o'clock headlines about ourselves.

Headphones make it worse. Podcast listeners wear them, which means every mouth click, every shaky breath, every little wobble lands right in their ears. I once recorded an intro five times. The words were fine every time. The energy was wrong — it sounded like someone reading a warranty, not a host welcoming you to a show.

What a phone greeting actually needs

So, back to my cousin's flower shop. An IVR greeting is neither of those things. It sits in the middle with its own personality.

Think about who's calling. Maybe they want to know if their order went out. Maybe they're annoyed because the roses arrived wilted. A too-hyped voice reads as fake when someone's upset, and a flat one sounds like you can't be bothered. Neutral-warm is the target, and it's a narrow target.

The bigger issue is repeat listenability. That same prompt might play four hundred times a day. A voice that's even slightly grating becomes torture by Tuesday, for your customers and for whoever sits closest to the phone.

What finally worked for her: slow the pace about ten percent below conversational, and leave real gaps between menu options. "For orders, press one. [beat] To speak with someone, press two." That beat is where the caller's brain catches up. Machine-gun delivery just manufactures pressed-the-wrong-option misery for everyone.

My home "studio" was its own disaster

My first attempt at her greeting, I just opened the laptop and talked. The recording had a fan whirring through every second and — I'm not joking — the neighbor's leaf blower soloing underneath. I deleted it out of pure embarrassment.

A few things I got wrong, in case you're about to do the same:

  • Mic distance. Too close and every "p" and "b" pops like a firecracker. Too far and the room takes over: echo, HVAC hum, that hollow sound. A hand's width away, slightly off to the side of the mouth, is the sweet spot.
  • The room. IVR audio gets compressed again by the phone system, so a little room noise that's invisible in a video becomes obvious fuzz after that. A closet full of coats beats an empty kitchen every time.
  • The recording app. Lots of free apps default to aggressive compression and shave off the high frequencies, and the voice comes out flat and lifeless. Record raw WAV if the app allows it. You can't add back what was never captured. Learned that one the expensive way.

You don't need a studio. You need a quiet room, a hanging coat, and a raw file.

Don't reuse your video voice on the phone

A friend who does product videos asked if he could splice his old narration clips into phone prompts. I tried it for him. It fell apart immediately.

Video narration performs. The pitch lifts on the exciting part, pauses stretch dramatically before the selling point. Great on camera. On a phone line, which throws away most of the frequency range anyway, all that acting turns into noise and the actual information — "press two for billing" — drowns in it.

An IVR voice is a receptionist who never sleeps, never gets tired, and never has a bad morning. It needs to be understood, not admired. Different brief, different recording.

Which is exactly why I stopped trying to record it myself.

The workflow I actually use now

I'm not a voice actor, I wasn't renting studio time for a flower shop, and outsourcing a forty-second prompt costs more than you'd guess. So here's what I do instead.

Start with about thirty seconds of clean speech in the style you want: calm, clear, friendly. Record it yourself or grab any decent sample. Upload it on the free voice cloning page and you can clone your voice free — no account, no card, no trial timer ticking in the corner. A couple of minutes later that voice is saved and reusable.

Then open the free text to speech page, paste in the script — "Thank you for calling Rose & Stem. For order status, press one. To talk with our team, press two." — generate, download. It's a no registration text to speech setup, the output carries no watermark, and it's unlimited free TTS, which matters because I regenerate a single line eleven times until the pause feels right. (I am that person.) Drop the file into whatever phone system you run — RingCentral, Grasshopper, a Twilio flow, even an old PBX — and the whole thing is done in an afternoon.

I'd already burned hours on sites that advertise online voice cloning free and then demand a card number at the download button, and half the free TTS online results I clicked wanted an account before playing a single second of audio. So this part was a relief. If you want to go further, there's a free voice clone API, with docs on the tutorials page, so you can generate prompts from your own tools instead of typing them by hand.

One honest caveat, because reviews that only gush are useless: a free AI voice generator won't rescue a bad script. The most realistic AI voice online still can't make "Press one to be connected to our customer satisfaction department" sound human, because nobody has ever said that sentence out loud in their life. Write like you talk. Short sentences. Real gaps between options. Free speech synthesis has come a long way from the GPS-lady era, and the ceiling now is your writing, not the robot. Text to speech online free tools are everywhere these days — the differentiator is the script and the sample you feed them.

FAQ

Can I use a cloned voice for my business phone greeting?

Yes. As long as the voice is clear and the script is easy to follow, it works fine for inbound prompts. One thing worth knowing if you're in the US: inbound greetings are one thing, but outbound automated calls are regulated, and the FCC has ruled that AI-generated voices in robocalls violate existing telemarketing rules. Greeting people who called you is fine. Cold-calling them with a cloned voice is a different legal world entirely.

Is audio from a free TTS tool okay to use commercially?

With this one, yes. The generated audio is watermark-free and fine for commercial use — phone systems, videos, podcasts, all of it. That said, check the terms on whatever tool you end up using, because "free" sites vary wildly on this and some quietly claim rights over your output.

Try it tonight: record thirty seconds of yourself saying your business greeting, upload it, clone it, generate the prompt, and play it next to whatever's on your phone line right now. My cousin switched hers the same afternoon, and she says callers stopped asking "sorry, what was the option for orders?" within a week. Small win. But you can hear it.