🎙️ Give Your Assistant a Voice
Transform text into natural, human-like speech with our integrated TTS providers. Each provider offers unique advantages for different use cases.
Quick Provider Comparison
⚡ ElevenLabs
Premium Quality & Customization70+ languages, voice cloning, advanced controlsBest for: High-quality customer interactions
🚀 Deepgram Aura
Ultra-Low Latency3x faster than competitors, phone-optimizedBest for: Real-time conversations
🎯 Cartesia Sonic 3
Multilingual Excellence42 languages, voice cloning, low latencyBest for: Multilingual agents, global deployments
☁️ Azure Speech
Enterprise Scale500+ neural voices, 100+ languagesBest for: Enterprise, Azure ecosystem
🎭 Inworld.ai
AI-Powered EmotionsMultilingual, emotional markup, voice cloningBest for: Expressive, contextual responses
🎙️ Resemble AI
Custom Voice CreationWebSocket streaming, personalized voicesBest for: Brand-specific voice identity
🧠 OpenAI TTS
Instruction-Aware Voices
tts-1, tts-1-hd, and gpt-4o-mini-ttsBest for: OpenAI-native stacks and voice instructions🧩 Additional Providers
Kokoro, Uplift, Murf, SonioxSelf-hosted and specialized provider optionsBest for: Custom deployments and specialized voices
Feature Matrix
| Provider | Latency | Languages | Voice Cloning | Streaming | Best For |
|---|---|---|---|---|---|
| ElevenLabs | Vendor-reported low latency | 70+ | ✅ Advanced | WebSocket | Premium quality |
| Deepgram | Vendor-reported very low latency | English | ❌ | WebSocket | Speed & phone calls |
| Cartesia | Vendor-reported low latency | 42 | ✅ Yes | WebSocket | Multilingual |
| Azure | Provider/config dependent | 100+ | ❌ | HTTP | Microsoft ecosystem |
| Inworld | Provider/config dependent | 11 | ✅ Zero-shot | HTTP/WS | Emotional expression |
| Resemble | Provider/config dependent | English | ✅ Custom | WebSocket | Brand voices |
| OpenAI | Provider/config dependent | English-first | ❌ | HTTP streaming | OpenAI-native stacks |
| Kokoro | Self-hosted dependent | Model-dependent | ❌ | HTTP | Self-hosted/local deployments |
| Uplift | Provider-dependent | Specialized | ❌ | Streaming service | Uplift voices |
| Murf | Provider-dependent | Multi-language | ❌ | HTTP | Style/rate/pitch controls |
| Soniox | Provider-dependent | Model-dependent | ❌ | Streaming service | Soniox voice stack |
New to TTS? Start with ElevenLabs for the best balance of quality and features, Deepgram if speed is your priority, or Cartesia for multilingual support.
Setup Overview
All providers follow the same basic setup pattern:Provider Deep Dives
🎭 ElevenLabs - Premium Voice Quality
Latest Models: Flash v2.5 (75ms), v3 (70+ languages), Turbo v2.5Key Features: Advanced voice controls, multilingual support, custom voice creationPerfect For: Customer service, content creation, multilingual applications→ Complete ElevenLabs Guide
⚡ Deepgram Aura - Ultra-Fast TTS
Latest Models: Aura-2 (next-gen), Aura (proven)Key Features: Industry-leading speed, phone optimization, µ-law encodingPerfect For: Real-time phone calls, live chat, interactive applications→ Complete Deepgram Guide
🎯 Cartesia Sonic 3 - Multilingual Excellence
Latest Models: sonic-3 (latest), sonic-3-2025-10-27 (stable)Key Features: 42 languages, voice cloning from ~5 sec, low latency WebSocketPerfect For: Global deployments, multilingual voice agents→ Complete Cartesia Guide
☁️ Azure Speech - Enterprise Scale
Latest Models: Neural (high-quality), Standard (basic)Key Features: 500+ voices, 100+ languages, SSML support, Microsoft integrationPerfect For: Enterprise applications, Azure ecosystem users→ Complete Azure Guide
🎪 Inworld.ai - AI-Powered Expression
Latest Models:
inworld-tts-1, inworld-tts-1-max, inworld-tts-1.5-max, inworld-tts-1.5-miniKey Features: Emotional markup, context awareness, 11 languagesPerfect For: Gaming, entertainment, emotional customer support→ Complete Inworld Guide🎙️ Resemble AI - Custom Brand Voices
Key Features: WebSocket streaming, unlimited voice creation, business plansPerfect For: Brand consistency, personalized experiences, enterprise→ Complete Resemble Guide
🧠 OpenAI TTS - Instruction-Aware Speech
Latest Models:
tts-1, tts-1-hd, gpt-4o-mini-ttsKey Features: Built-in OpenAI voices, speed control, instruction support on GPT-4o mini TTS modelsPerfect For: Teams already using OpenAI keys and model-specific voice instructions→ Complete OpenAI TTS GuideAdditional Supported Providers
The backend also supports these TTS providers throughtts_settings.provider:
| Provider | Provider Key | Important Options |
|---|---|---|
| Kokoro | kokoro | voice_id, model_id, language, speed, kokoro_base_url; no API key required by default |
| Uplift | uplift | voice_id, model_id, language, output_format |
| Murf | murf | voice_id, model_id, style, rate, pitch, region, optional variation, locale, pronunciation_dictionary |
| Soniox | soniox | voice_id, model_id, language |
AWS Polly and Google TTS appear only as future placeholders in backend enum comments and are not registered in the active TTS factory. Do not document them as supported providers until they are wired.
Advanced Topics
🎛️ Voice Tuning
Master stability, similarity, and style controls across all providers
🔧 Troubleshooting
Common issues and solutions with step-by-step fixes
📈 Best Practices
Performance optimization, cost reduction, and production tips
Related Documentation
🔗 See Also
Configuration: Learn how to configure TTS in your AI Configuration settings.Integration: Understand how TTS fits into the overall Architecture of Burki Voice AI.Call Management: Discover how TTS works with Call Management features.
Quick Start Guide
- Business Calls
- Customer Support
- Multilingual
- Enterprise
- Real-time Apps
Recommended Setup:
- Provider: ElevenLabs or Deepgram
- Model: Flash v2.5 or Aura-2
- Voice: Professional (Rachel, Asteria)
- Settings: Stability 0.5, Speaker Boost ON
API Rate Limits: Each provider has different rate limits and pricing models. Check the individual provider pages for detailed pricing information.
🚀 Ready to Get Started?
Choose your provider and dive into the detailed setup guides, or check out our Best Practices for optimization tips.