AI voice notes: your setter replies in your real voice

Text is fast, but a voice note lands. SetterStack can send real voice notes in your Instagram DMs — spoken in your own voice — so a lead hears you, not a wall of text. You record about a minute once; from then on your AI decides when a spoken reply will hit harder than typing, and sends one automatically.

Instagram only for now. Voice notes currently send on Instagram. On other platforms your AI replies with text as usual — voice for more platforms is coming.

This isn't a robot voice reading a script. It's you: your tone, your pacing, your warmth — at the speed of automation.

01Why voice notes convert

A voice note feels one-to-one in a way text can't fake. It reads as high-effort and personal — the kind of thing a busy person only does for someone they actually want to talk to. So it earns trust and replies:

• It feels human and personal. Hearing a real voice at the right moment breaks through a crowded inbox and makes the lead feel seen.
• It carries emotion. Reassurance, excitement, and empathy come through in a voice in a way they never do in text — perfect for warming someone up or answering a nervous objection.
• It signals effort. A voice note says "a real person cares about this conversation," which lifts trust and response rates.

You stay honest the whole time — it's genuinely your voice, sent by your assistant on your behalf.

02When your AI sends one — you set the frequency

You control how eagerly your AI reaches for a voice note with a How often slider — five steps from "never sends first" to "voice-first":

• Only when they voice — never sends a voice note on its own; only replies in voice if the lead sends one.
• Rarely — mostly text; a voice note only for big moments (a warm welcome, a nervous objection, the moment it locks the call).
• Sometimes — a balanced mix; voice when it adds warmth or energy, text otherwise.
• Often — leans into voice; sends spoken replies regularly through the conversation.
• Almost always — voice-first; nearly every reply is a voice note (text only for links, prices, or lists).

Slide to wherever fits your style, and you'll see a plain-English description of exactly what that setting does. No matter where you land, if the lead sends you a voice note first, your AI replies with one back.

Sends several voice notes in a row? If they come close together, your setter picks them all up and replies once to the whole thing — instead of firing off a separate reply to each note before the lead has finished their thought. Length varies naturally — a quick hello stays short, a thoughtful answer runs longer — and if anything ever prevents a voice note from sending, the AI automatically falls back to text, so the conversation never stalls — and that fallback is recorded, so a voice problem shows up as a problem instead of quietly becoming text forever.

03How your voice sounds

Three dials in the Voice tab shape the delivery of every spoken note — separate from your typed style, so your voice can feel one way and your texts another:

• Voice energy — from Steady (calm, even) to Expressive (lively, more range). Where you land changes how animated your setter sounds.
• Voice note tone — a short description of the vibe you want spoken ("warm and unhurried," "sharp and upbeat"). Leave it blank to just sound like your recording.
• Voice naturalness — how many natural pauses and filler words ("uh," "you know," a beat before answering) get woven in, so it sounds spoken-in-the-moment instead of read off a page.
• Accent strength — how tightly the voice sticks to your recording. Higher keeps a distinct or strong accent intact; lower lets it smooth toward a more generic sound. If your accent is getting neutralized, push this up — and for the biggest effect, re-record in a quiet spot speaking in your natural accent (don't try to "flatten" it).

Adjust a dial, then tap "Hear a sample" — the preview reflects your change right away, no need to Save first. Same in the Test playground: the AI's reply uses whatever you've currently got set on screen, so you can tune and hear the difference immediately.

04Backgrounds: a voice note from wherever they are

A voice note feels most real when it sounds like it was fired off between things — from the car, the gym, a busy terminal. In the Voice tab you can turn on background sounds so your setter feels like an actual person moving through their day, not a bot in a silent booth.

Pick the places your setter can be — Driving, Gym, Airport, Office — and your AI does the rest:

• One place at a time. Your setter stays in a single spot for a couple of hours, then moves on to the next — so everyone who hears a voice note around the same time hears the same place, exactly like a real person. The routine shifts a little day to day, so it never feels scripted.
• Evenings wind down. Later on, it settles into a quiet, at-home sound.
• Nights go quiet. In the middle of the night it's a faint, in-bed vibe — and your AI rarely sends a voice note at that hour anyway, because real people are asleep. Late-night messages almost always get a text back instead.

When a background is on, the first voice note in a conversation opens with a quick, breezy mention of where they are — "driving right now but," "at the gym but," — then gets straight to the point. It's said once, casually, not on every message.

Hear any of them before you commit. Tap Hear it on a scene to play a sample in your own voice over that background. Prefer your own recording? Upload your own clip on any scene (a short 10–30 second loop is plenty) and it'll use that instead.

Add your own places. Beyond the built-in scenes, you can create your own — upload a clip, name the place (the beach, my studio, the shop), and your AI will say that name in the opener ("at the beach but,") and fold it into the daily rotation. Everyday places only; anything inappropriate is rejected.

Backgrounds are optional — leave them all off and your voice notes send clean. They're a finishing touch that makes an already-personal voice note feel unmistakably human.

05Which languages your voice speaks

Your AI already replies in whatever language the lead writes in — that's great for text. But a spoken voice note is different: if your AI speaks a language you can't, the lead shows up to the call expecting it, and you're stuck.

So in the Voice tab you tell SetterStack which languages you can actually hold a call in. Then:

• If a lead writes in a language you speak → the voice note is spoken in their language.
• If a lead writes in a language you don't speak → your text reply still matches their language, but the voice note is spoken in your primary language and warmly lets them know you don't speak theirs (so there's no awkward surprise on the call).

Pick every language you're comfortable speaking; the first one you add is your primary (fallback) language. Leave it empty only if you want the voice to mirror any language — not recommended, since it can promise a call in a language you can't deliver.

06Voice identity: accent, gender and age

Under your languages you'll find Voice identity — where your accent is from, Gender and Age.

Your accent doesn't have to match the language you speak. Plenty of people speak English with a Russian, French or Brazilian accent, so you pick two things:

• Accent is from — the language your accent comes from (say, Russian), even if you're speaking English.
• Which accent — the region within it (Moscow, Saint Petersburg, Standard).

Speak English with an American accent? Set both to English → American. Speak English with a Moscow accent? Set English as your language and Russian → Moscow as your accent.

Important: this describes your accent, it doesn't create one. Setting a French accent on a voice that doesn't have one won't make it sound French — your accent comes from the audio you recorded, and nothing here can add one that isn't in it. (If anything, turning Accent strength up makes the voice sound more like your original recording, not less.) The only way to get a different accent is for someone with that accent to record it.

These labels keep your voice properly identified, and they're a useful prompt to get your accent right. Worth knowing what actually shapes how your accent sounds, in order of impact:

1. Your recording. The accent comes from the audio you read out. Speak in your natural accent — don't flatten it or "neutralize" it for the microphone. This matters more than every setting combined.
2. Your primary language. The first language in your list anchors pronunciation, which is what keeps a strong accent from drifting.
3. Accent strength. The dial above. On a distinct regional accent, set it to Strong — a looser setting lets the voice drift toward a generic, placeless sound.

If your accent came out weaker than you expected, re-record rather than hunting through settings. Your voice is captured at the moment you record it, so a fresh recording in your natural accent is the real fix.

07Setting it up

Everything voice lives in its own Voice tab now (right next to Personality), so it's easy to find and Personality stays uncluttered. In AI Config → Voice:

Setup runs in three steps:

1. Record audio. Read the short script out loud for about 30–60 seconds, in your natural accent. Don't flatten it — this recording is where your accent comes from.
2. Voice information. Name the voice, say where your accent is from, and describe it in your own words ("Nico, who has a strong french accent"). Add gender and age if you like.
3. Try it out. Type anything and hear it back in your voice — exactly what a lead would receive. Tweak and re-generate until it sounds like you.

Then set how often and pick your languages (see above), flip voice notes on, and save. From then on your AI sends spoken replies at the right moments.

That's it — no editing, no re-recording per message. Record once and your setter can speak for the rest.

08Tips for the best-sounding voice

• Record in your natural, relaxed tone — the way you'd actually talk to a lead, not a formal "podcast voice." Your leads should recognize you.
• A longer sample makes a truer voice. Reading the full script (closer to 60 seconds than 15) captures more of your range, so the voice notes sound more like the real you.
• Find a quiet spot. Less background noise means a cleaner, more natural result.
• Preview before you go live. Use Hear a sample until it sounds like you, then enable it.
• Watch for the microphone warning. If your browser is processing the mic, you'll see an amber note above the script before you record. The big one is auto gain control: it evens out your volume, which flattens the natural rise and fall in how you talk — and a voice recorded through it usually comes back sounding low and a bit lifeless. You can still record, but Chrome on a laptop (or a different mic) will give you a noticeably better clone.

09"It doesn't sound right" — just say so

Under AI → Voice there's a box that asks How did that sound? Tell it what's off in your own words — "it sounds too robotic", "too excited", "it doesn't sound like me", "too polished, like reading", "too many ums" — and it adjusts the one setting that actually controls it.

Nothing changes until you say so. You'll see exactly what it wants to change (for example Voice energy: Balanced → Lively), you can Hear it first, then Keep it or Discard. There's also Reset to recommended, which puts every voice dial back to our tuned default in one tap.

Two things worth knowing:

• Some problems aren't a settings problem. If your accent is getting smoothed out, or it still sounds flat at the most expressive setting, it'll tell you that a fresh recording will do more than any dial — a cloned voice can only keep what's in your audio, it can never add it.
• Pace isn't a dial. How fast you talk comes from your Voice note tone description. Write it the way you actually speak and the delivery follows.

10Voice is set per profile, never shared

Your cloned voice is yours — it lives on your profile, not on the whole team. If you run multiple profiles (an agency or a team), each person records and enables their own voice inside their own profile; one person's voice can never become the default for everyone. That's why voice isn't editable from the "All Profiles" team view — switch to a specific profile to set it. (Voice notes currently send on Instagram; other platforms reply in text, with more coming.)

11Text still does the heavy lifting

Voice notes are a spice, not the whole meal. Most replies stay text — fast, skimmable, and easy to act on. Voice notes are the moments that deserve a human touch, layered on top of everything else your AI already does: qualifying the lead, handling objections, and driving to your conversion goal — whether that's a booked call, a free trial, your website, a webinar, or a visit to your location.