Skip to main content

🎭 Inworld.ai: AI-Powered Expression

Advanced AI-driven TTS with emotional markup, context awareness, and support for 11 languages. Perfect for gaming, entertainment, and expressive customer interactions.

​
Quick Setup

1

Get API Key

  1. Visit Inworld Studio and create an account
  2. Navigate to Integrations → API Keys
  3. Generate a new API key for TTS
  4. Copy your Bearer token
2

Configure in Burki

  1. Go to AI Configuration → TTS tab
  2. Select Inworld.ai as provider
  3. Paste your Bearer token in the TTS API Key field
3

Choose Model & Voice

Select TTS-1 or TTS-1-Max model and your preferred voice

​
Available Models

🎪 TTS-1

~200ms latencyFlagship model with realistic, context-aware synthesisLanguages: 11 supported languages Best for: Production applications, customer service

🔬 TTS-1-Max

~250ms latencyLarger, more expressive model (experimental)Languages: 11 supported languages Best for: Creative content, gaming, entertainment

​
Multilingual Voice Library

  • English
  • Spanish
  • French
  • Other Languages

Hades

Deep and commandingPerfect for authoritative charactersVoice ID: Hades

Alex

Clear and naturalGreat for professional applicationsVoice ID: Alex

Ashley

Warm and friendlyIdeal for customer serviceVoice ID: Ashley

Aria

Professional and articulatePerfect for business communicationsVoice ID: Aria

​
Emotional Markup System

🎭 Express Emotions in Speech

Inworld’s unique emotional markup allows you to add feelings and speaking styles directly in your text.

​
Emotional Tags

  • Basic Emotions
  • Speaking Styles
  • Advanced Examples
[happy] I'm excited to help you today!
[sad] I'm sorry to hear about that issue.
[angry] This is completely unacceptable!
[surprised] Wow, I didn't expect that!
[fearful] Please be careful with that.
[disgusted] That's not what I ordered.
Available emotions: happy, sad, angry, surprised, fearful, disgusted, neutral

​
Zero-Shot Voice Cloning

🎯 Custom Voice Creation

Create custom voices without training data. Just provide a voice ID and Inworld handles the rest.
1

Enable Custom Voice

In your assistant configuration, select “Custom” voice option
2

Enter Voice ID

Provide your custom voice identifier in the Custom Voice ID field
3

Test & Refine

Test with sample text and adjust based on results

​
Language Support & Quality

🌍 Production-Ready Languages

Inworld provides native-quality voices for multiple languages with varying production readiness.
LanguageStatusVoices AvailableQuality Rating
English🟢 Production4+ voices⭐⭐⭐⭐⭐
Spanish🟢 Production4+ voices⭐⭐⭐⭐⭐
French🟢 Production4+ voices⭐⭐⭐⭐
German🟡 Beta2+ voices⭐⭐⭐⭐
Chinese🟡 Beta4+ voices⭐⭐⭐⭐
Japanese🟡 Beta2+ voices⭐⭐⭐
Italian🟡 Beta2+ voices⭐⭐⭐
Portuguese🟡 Beta2+ voices⭐⭐⭐
Dutch🟡 Beta2+ voices⭐⭐⭐
Korean🟡 Beta2+ voices⭐⭐⭐
Polish🟡 Beta2+ voices⭐⭐⭐

​
API Integration

import requests
import json

def synthesize_with_emotions(text, voice_id="Ashley", language="en"):
    url = "https://api.inworld.ai/v1/text-to-speech/stream"
    
    headers = {
        "Authorization": "Bearer YOUR_INWORLD_API_KEY",
        "Content-Type": "application/json"
    }
    
    data = {
        "text": text,
        "voice_id": voice_id,
        "model": "inworld-tts-1",
        "language": language,
        "output_format": "wav",
        "sample_rate": 8000
    }
    
    response = requests.post(url, json=data, headers=headers, stream=True)
    return response

# Example with emotional markup
text_with_emotion = "[happy] Welcome to our service! [excited] How can I help you today?"
response = synthesize_with_emotions(text_with_emotion, "Ashley", "en")

​
Use Case Examples

  • Customer Support
  • Gaming/Entertainment
  • Educational Content
  • Multilingual Business
# Empathetic customer service
[understanding] I completely understand your frustration. [reassuring] Let me help you resolve this issue right away. [confident] I'll have this sorted out for you in just a moment.
Best voices: Ashley (EN), Lupita (ES), Hélène (FR) Model: TTS-1 for consistency

​
Configuration Examples

🎛️ Optimal Settings

Recommended configurations for different use cases.
{
  "model": "inworld-tts-1",
  "voice_id": "Ashley",
  "language": "en",
  "output_format": "wav",
  "sample_rate": 8000,
  "text": "[friendly] Thank you for calling! [helpful] How may I assist you today?"
}
Features:
  • Phone-compatible audio format
  • Friendly, professional tone
  • Emotional warmth in greeting
{
  "model": "inworld-tts-1",
  "voice_id": "Diego",
  "language": "es",
  "output_format": "wav",
  "sample_rate": 16000,
  "text": "[amigable] ¡Bienvenido! [servicial] ¿En qué puedo ayudarte hoy?"
}
Features:
  • Native Spanish voice
  • Cultural appropriate emotions
  • Higher quality audio format
{
  "model": "inworld-tts-1-max",
  "voice_id": "Hades",
  "language": "en",
  "output_format": "wav",
  "sample_rate": 24000,
  "text": "[menacing] You dare enter my domain? [angry] Prepare to face my wrath!"
}
Features:
  • Maximum expression model
  • Character-appropriate voice
  • High-quality audio for immersion

​
Best Practices

🎯 Maximize Inworld's Emotional AI

Get the most out of Inworld’s unique features with these proven strategies.
Do’s:
  • Use emotions that match the content context
  • Place tags at natural speech boundaries
  • Mix emotions sparingly for realistic conversations
  • Test different voices with the same emotions
Don’ts:
  • Overuse emotional tags (max 2-3 per sentence)
  • Use conflicting emotions close together
  • Rely only on markup - the text itself should convey meaning
  • Mix languages and emotions in complex ways
For Business Applications:
  • English: Ashley, Alex, Aria
  • Spanish: Lupita, Rafael
  • French: Hélène, Mathieu
For Creative Content:
  • Dramatic: Hades, Aria
  • Friendly: Ashley, Diego
  • Professional: Alex, Johanna
For Gaming:
  • Heroes: Aria, Alex
  • Villains: Hades
  • NPCs: Ashley, Diego
# Efficient text processing
def optimize_for_inworld(text):
    # Keep sentences under 50 words for best latency
    sentences = split_into_sentences(text)
    optimized = []
    
    for sentence in sentences:
        if len(sentence.split()) > 50:
            # Break long sentences at natural points
            optimized.extend(break_long_sentence(sentence))
        else:
            optimized.append(sentence)
    
    return optimized

# Cache voice models for repeated use
voice_cache = {}

def get_cached_voice(voice_id, language):
    cache_key = f"{voice_id}_{language}"
    if cache_key not in voice_cache:
        voice_cache[cache_key] = load_voice_model(voice_id, language)
    return voice_cache[cache_key]

​
Pricing

💰 Flexible Pricing Model

Pay-per-character with volume discounts and free tier for testing.
PlanCharacters/MonthPriceFeatures
Free25,000$0All voices, emotional markup
Starter100,000$9Priority processing
Professional500,000$29Custom voice support
EnterpriseCustomCustomDedicated support, SLA

​
Troubleshooting

Problem: Emotions don’t seem to affect the voiceSolutions:
  • Verify emotion tag spelling (case-sensitive)
  • Check if the voice supports the specific emotion
  • Try with TTS-1-Max model for stronger expression
  • Ensure tags are properly formatted: [emotion]
Problem: Wrong language pronunciationSolutions:
  • Explicitly set language parameter in API call
  • Use native voices for each language
  • Avoid mixing languages in single requests
  • Verify voice ID supports the target language
Problem: Custom voice ID not workingSolutions:
  • Verify custom voice ID with Inworld support
  • Check if voice is properly trained and available
  • Use fallback to standard voices during setup
  • Contact support for voice activation status

​
Migration Guide

  • From Other Providers
  • Voice Mapping
Key Advantages of Switching:
  • Emotional markup for better user engagement
  • Native multilingual support (11 languages)
  • Zero-shot voice cloning capabilities
  • AI-powered context awareness
Migration Steps:
  1. Map existing voice preferences to Inworld voices
  2. Add emotional markup to enhance user experience
  3. Test multilingual capabilities if applicable
  4. Optimize for Inworld’s strengths (emotions, languages)

🎭 Ready to Add Emotion to Your AI?

Configure Inworld.ai in your assistant settings and start creating more engaging, expressive conversations!