How I make AI voices sound human

The exact voice settings behind my skits. Three tricks: emotion tags, low stability, one line per file.

The recording booth

<aside> 🔄

Use any AI voice tool you like. I used ElevenLabs v3. Any tool works, as long as it sounds natural and lets you control the emotion of each line.

</aside>

Before you start: pick 2 voices

Pick two voices that sound clearly different, so viewers always know who's talking.

Me: a warm, younger voice (I used Eric, speed 1.0)
My AI sidekick: a dry, slightly British voice (I used Charlie, speed 0.9)

Trick 1: Emotion tags

Write the feeling in square brackets at the start of the line. The voice acts it out.

[deadpan] It is.
[whispering] Like this?
[softly, sincere] Honestly, it sounds like you on a good day.
[mock-solemn] And yes, my voice is fake too.
[warm, friendly] Comment VOICE and I'll send you the settings.

Tags I use a lot:

Funny	Soft	Energy
[deadpan]	[softly, sincere]	[confident, explaining]
[sarcastic]	[touched, surprised]	[smug, proud]
[mock-solemn]	[whispering]	[exasperated]
[dry, annoyed]	[worried, confiding]	[curious]

<aside> ✅

Two words beat one. "[softly, sincere]" gives a better read than just "[sad]".

</aside>

Trick 2: Stability around 0.3

Stability controls how much the voice changes from line to line.

Low (about 0.3): more emotion, more human. This is what I use for every line.
High (1.0): flat and robotic. Every line sounds the same.