Text DMs get ignored because they read like a template nobody bothered to hide. Voice notes change the dynamic, and AI lets you send them without recording each one by hand. This page covers how to set it up and where it goes wrong.
The short version
Going deeper
Most LinkedIn outreach reads like a template with a first name swapped in. People can smell it in the first line and they keep scrolling.
The problem is not that text is bad. The problem is that everyone defaults to it, so your message lands in a pile of identical ones.
A voice note breaks that pattern. It is rare enough that people actually open it, and the human voice carries warmth that no amount of clever copy can fake.
ElevenLabs clones your voice from a short sample, then generates new audio from text you type. So you write the message and it speaks in your voice.
That means you can produce dozens of personalized voice notes without sitting in front of a mic for an hour. The first line references something real about the prospect, the rest carries your usual pitch.
lemlist reported doubling reply rates by automating personalized voice messaging with this approach. My read is that the lift comes less from the tech and more from the fact that so few people are doing it at all.
A voice note with no personalization is just a text DM that takes longer to listen to. The format alone does not save you.
The move is to pull one specific detail per prospect - a recent post, a role change, a shared connection - and open with it. Tools that connect a lead sheet to AI research and then to ElevenLabs can do this in a pipeline, so the research and the audio generation happen in one pass.
Without that detail, you are scaling something that should not be scaled.
There is a real tension here. The reel says to make sure the other person has no idea the voice is AI generated, and that works for getting replies. But it sits in murky territory.
I think the cleaner version is using AI to scale your actual voice and your actual words, not to pretend you spent time you did not spend. The voice is yours. The message is something you genuinely would say.
That distinction matters more as these tools spread. The first time someone realizes a voice note was synthetic and feels tricked, your reputation takes the hit, not the software's.
Start small before you automate anything. Record a clean voice sample for ElevenLabs, write five real outreach messages by hand, and generate them as audio. Listen back and check that it sounds like you on a normal day, not a presenter reading a script.
Once the voice is right, connect a lead source like a spreadsheet, add a research step so each message has one specific hook, and route the audio out through your sending tool.
Keep the volume low enough that you can read every message before it goes. Scale is useless if the quality slips and the replies dry up.
Want to go deeper?
One call, one on one. We take this and apply it to your account or your workflow.
Book a callSources
