Emotion is heard. Emotion is what makes a voice stick with you, the charm of a 1920s flapper, the grit of a hard-boiled detective, the warmth of a bedtime storyteller. These moments of character are what make conversations feel human.
Introducing character datasets, curated collections of voices performing over 450 characters with personality, style, and emotion. These recordings teach your models to go beyond robotic speech and deliver dialogue with nuance, energy, and connection.
Don’t Take Our Word for It. Listen for Yourself!
Wise Wizard
1920s Flapper
Tone: Witty
Emotion: Amusement
Weary Time Traveller
Tone: Dry
Emotion: Boredom
Tone: Inspirational
Emotion: Joy
Head Chef of a Michelin Star Restaurant
Tone: Authoritative
Emotion: Disgust
Lazy Housecat (as a person)
Tone: Sarcastic
Emotion: Amusement
Drill Sergeant
Tone: Authoritative
Emotion: Pride
What’s Inside
Hundreds of Hours of Fine-Tuning Data – 600+ hours available now, with 700 more added every month.
Professional Voice Talent – authentic, English language recordings across diverse characters, tones, and styles.
Extensive Character Set - 450+ tagged character types ready to train and test.
Ready-to-Use Metadata – delivered in JSON with text, character, and emotion tags.
Flexible Formats – audio in .wav (or preferred format) for seamless ingestion.
Rich JSON and Metadata
Get everything you need to ingest, train, and audit at scale—without the guesswork. The Characters Dataset includes clean, consistent JSON per asset.
Sample JSON
{
"script": {
"title": "Sample Script 7-7",
"content": "Wait a minute. 4... 8... 15... 16... 23... hold on. Let me get my glasses. 4... 8... 15... 16... 23... 42. [gasp] No. No, it can't be. This is a joke, right? I've got... I've got all of them. Every single one. I... I won? I WON? [laugh] I WON THE LOTTERY! I'M RICH! I can't believe it! I almost used this ticket as a bookmark! This is... this changes everything! Oh my god!",
"tone": "informal",
"emotion": "surprise (positive)",
"character_type": "Lottery Winner",
"direction": "Checking the winning numbers on a ticket they almost threw away."
},
"length": 57589
}
Why Voices?
Ethically Sourced – contributor consent and clear licensing for risk-free use.
Scale You Can Trust – 4M+ global voice contributors for unmatched diversity.
Proven Partner – trusted by the world’s leading technology companies, including Meta, Adobe, and SoundHound.
Connect With Us to Learn More
Be among the first to access this one-of-a-kind dataset and supercharge your AI model training with data that goes far beyond generic speech.
