AI voice messages work best when they feel conversational, like you're speaking to a real person. Follow our simple tips to make your messages more human.
💡 If you're new to AI voice messages, see our LinkedIn AI voice messages article for an overview of how to create and send them.
Add pauses
Use “…” or “–” to create short, natural breaks. This helps the message flow like real speech.
Use light filler words
Words like “umm,” “so,” “well,” or “you know” make the tone relaxed and friendly. Just use them sparingly.
Use informal contractions
Say “gonna,” “wanna,” “kinda,” or “gotta” to sound less scripted.
Write spoken contractions naturally
Phrases like “I'd like to,” “don't know,” or “we're going to” read more like real speech.
Spell out symbols and abbreviations
Write words instead of characters:
“%” → percent
“$” → dollars
“TB” → terabyte
“m” → million
“k” → thousand
Don't use links
Links don't sound natural when read aloud and can break the flow.
Keep sentences short
Break complex ideas into clear, simple sentences for easier listening.
Write like you talk
Use a friendly, informal tone – imagine speaking directly to someone instead of writing an email.
Use personalization
Add variables like {{FirstName}} to make your message feel personal and genuine.
Keep it short
Short messages are more engaging. Aim for 20-40 seconds (LinkedIn allows up to 1 minute).
Use voice tags
Our voice model supports tags that help make speech sound more human and expressive.
To use them, place tags in lowercase square brackets at the start of a sentence or phrase you want them to apply to.
👉 Example:
[curious] Are you currently using any automation tools in your workflow?
You can also click Enhance in the editor. AI will automatically add voice tags throughout the message. You can review, edit, or remove them before saving.
The tags listed below are examples, not a fixed set. The model is flexible and can understand other voice-related words in brackets as well. You can experiment with different expressions depending on context.
Emotional expressions | These tags guide tone and attitude: [curious] [excited] [sad] [happy] [sarcastic] [dismissive] |
Vocal delivery & pacing | These control speed, volume, and delivery: [pauses] [slowly] [whispers] |
Non-speech sounds & reactions | These add natural human reactions: [exhales] [clears throat] [coughs] [sighs] [chuckles] [laughs] [frustrated sigh] [happy gasp] |
Add background noise
You can add ambient sounds, such as an office or restaurant environment, to make voice messages sound more natural.
💡 Background noise is not available for custom audio uploads.
Use the Pronunciation dictionary
The Pronunciation dictionary lets you control exactly how AI pronounces specific words in voice messages. This can be useful for company names, product names, industry terms, or other words that AI may mispronounce.
Pronunciation settings apply to all of your voice profiles and all newly generated AI voice messages.
To add a pronunciation rule:
Go to Settings → Voices → Pronunciation Dictionary.
Click Add pronunciation.
Enter the word and its pronunciation using International Phonetic Alphabet (IPA) notation.
Save your changes.
For example:
Jason → ˈdʒeɪsən
Reply → rɪˈplaɪ
💡 Important notes:
Add each word on a separate line. If you need to define the pronunciation of a phrase or multiple words, create a separate entry for each word.
Pronunciation rules apply only to AI-generated voice messages. They do not apply to uploaded audio files or pre-recorded voice messages.
Changes affect only newly generated voice messages. Previously generated or already sent messages will not be updated automatically.
The system does not automatically detect mispronounced words. You'll need to add custom pronunciations manually.