> ## Documentation Index > Fetch the complete documentation index at: https://docs.callintel.io/llms.txt > Use this file to discover all available pages before exploring further. # Text-to-Speech Integration > Configure OpenAI, ElevenLabs, Cartesia, Rime AI, and Inworld for voice output. Text-to-Speech providers convert agent responses into natural-sounding speech. This guide covers available providers and setup. ## Available Providers ### OpenAI (Recommended - Best Balance) **Why Choose OpenAI:** * Natural sounding voices * Wide language support * Good cost structure * Easy setup (if already using GPT) ``` Quality: Excellent Speed: 1-3 seconds to first audio Cost: ~$0.015 per 1K characters Languages: 50+ Voices: 5 options ``` ### ElevenLabs (Best Voice Quality) **Why Choose ElevenLabs:** * Most natural sounding voices * Extensive voice library * Voice cloning available * Premium sound quality ``` Quality: Premium Speed: 1-2 seconds Cost: ~$0.03 per 1K characters Languages: 30+ Voices: 100+ (including custom clones) ``` ### Cartesia (Enterprise Grade) **Why Choose Cartesia:** * Studio-quality audio * Ultra-low latency * Extreme customization * Enterprise support ``` Quality: Studio Speed: < 1 second Cost: Custom pricing Languages: 40+ Voices: Customizable ``` ### Rime AI (Fastest) **Why Choose Rime AI:** * Fastest synthesis * Good quality * Real-time performance * Lower cost ``` Quality: Good Speed: < 500ms (fastest) Cost: ~$0.01 per 1K characters Languages: 30+ Voices: 20+ ``` ### Inworld AI (Character Voices) **Why Choose Inworld:** * Character-specific voices * Interactive responses * Custom personality * Story-driven conversations ``` Quality: Good Speed: 1-2 seconds Cost: Custom pricing Languages: 20+ Voices: Character personas ``` ## Provider Comparison | Feature | OpenAI | ElevenLabs | Cartesia | Rime AI | | ------------- | ------ | ---------- | -------- | ------- | | Voice Quality | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | | Speed | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | | Cost | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐ | ⭐⭐⭐⭐⭐ | | Voices | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | | Languages | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ## Setup ### OpenAI TTS Setup (Recommended) **Step 1: Get API Key** If you're already using OpenAI for LLM, you have everything needed. 1. Go to [platform.openai.com/api-keys](https://platform.openai.com/api-keys) 2. Copy API key (or use existing) 3. Ensure TTS is enabled in account **Step 2: Add to CallIntel** 1. Go to Settings → Developer Settings 2. Click "API Keys" 3. Select "OpenAI" from provider list (if not already added) 4. Paste API key 5. Save **Step 3: Configure Agent** 1. Create/Edit Agent 2. Under "Text-to-Speech", select: * `tts-1` (fastest) * `tts-1-hd` (highest quality) 3. Choose voice: * Alloy, Echo, Fable, Onyx, Nova, Shimmer 4. Select language 5. Adjust speed (0.25 - 4.0) 6. Save **Step 4: Test** 1. Make a web call 2. Listen to voice quality 3. Adjust if needed ### ElevenLabs Setup **Step 1: Create Account** 1. Visit [elevenlabs.io](https://elevenlabs.io) 2. Sign up for account 3. Verify email **Step 2: Get API Key** 1. Go to Account Settings 2. Click "API Keys" 3. Copy your API Key 4. Save securely **Step 3: Add to CallIntel** 1. Go to Settings → Developer Settings 2. Click "API Keys" 3. Select "ElevenLabs" from provider list 4. Paste API key 5. Click "Test Connection" 6. Save **Step 4: Configure Agent** 1. Create/Edit Agent 2. Under "Text-to-Speech", select "ElevenLabs" 3. Choose voice: * Browse available voices * Listen to samples * Select favorite 4. Configure stability & clarity 5. Save **Step 5: Optional - Clone Your Voice** To create custom voice: 1. Go to ElevenLabs dashboard 2. Click "Voice Lab" 3. Select "Clone Voice" 4. Record 30-120 seconds of your voice 5. ElevenLabs creates voice 6. Use in agents ### Cartesia Setup **Step 1: Contact Sales** 1. Visit [cartesia.ai](https://cartesia.ai) 2. Request enterprise access 3. Complete onboarding 4. Receive API credentials **Step 2: Configure in CallIntel** 1. Go to Settings → Developer Settings 2. Add Cartesia API key 3. Configure for TTS 4. Save **Step 3: Configure Agent** 1. Create/Edit Agent 2. Select Cartesia for TTS 3. Choose voice (provided by sales) 4. Configure quality settings 5. Save ### Rime AI Setup **Step 1: Create Account** 1. Visit [rime.ai](https://rime.ai) 2. Sign up for account 3. Verify email **Step 2: Get API Key** 1. Go to Dashboard 2. Click "API Keys" 3. Create new key 4. Copy key **Step 3: Add to CallIntel** 1. Go to Settings → Developer Settings 2. Select "Rime AI" 3. Paste API key 4. Save **Step 4: Configure Agent** 1. Create/Edit Agent 2. Select "Rime AI" for TTS 3. Choose voice 4. Adjust speed and tone 5. Save ## Cost Comparison ### Monthly Costs (1000 calls, 2 min avg, 200 chars/min) ``` OpenAI (tts-1): 1000 calls × 2 min × 200 chars = 400,000 chars 400,000 × ($0.015/1000 chars) = $6/month Cost per call: $0.006 ElevenLabs: 1000 calls × 2 min × 200 chars = 400,000 chars 400,000 × ($0.03/1000 chars) = $12/month Cost per call: $0.012 Rime AI: 1000 calls × 2 min × 200 chars = 400,000 chars 400,000 × ($0.01/1000 chars) = $4/month Cost per call: $0.004 Cartesia: Custom pricing (typically enterprise rates) ``` **Cost Ranking:** 1. Rime AI (cheapest) 2. OpenAI (good value) 3. ElevenLabs (premium) 4. Cartesia (enterprise) ## Voice Selection ### OpenAI Voices | Voice | Gender | Style | Use Case | | ------- | --------- | ------------ | ------------------- | | Alloy | Neutral | Professional | Customer service | | Echo | Male | Clear | Sales, support | | Fable | Narrative | Storytelling | Interactive scripts | | Onyx | Male | Deep, warm | Executive calls | | Nova | Female | Professional | General purpose | | Shimmer | Female | Upbeat | Friendly service | **Recommendation:** * Customer service: Nova or Alloy * Sales: Echo * Executive: Onyx * Friendly tone: Shimmer ### ElevenLabs Voice Selection Browse 100+ voices by: ``` - Gender (male, female, neutral) - Age (young, middle-aged, senior) - Accent (American, British, etc.) - Tone (friendly, professional, etc.) ``` **Tips:** 1. Listen to samples 2. Test with your script 3. Consider context 4. Select voice that matches brand ### Custom Voice Cloning Create voice that sounds like you: ``` Step 1: Record yourself (30-120 seconds) Step 2: ElevenLabs processes audio Step 3: Clone available for use Step 4: Use in agents Cost: Included with subscription Quality: Premium Use: Brand consistency ``` ## Configuration Options ### Speed Control Adjust how fast agent speaks: ``` 0.5x = Very slow (for complex topics) 0.75x = Slow (for clarity) 1.0x = Normal (default) 1.25x = Fast (for efficiency) 1.5x = Very fast (for quick updates) Recommendation: 1.0x for customer service ``` ### Voice Quality **OpenAI Quality Options:** ``` tts-1: Faster, lower latency (50ms latency) tts-1-hd: Higher quality audio (100ms latency) Recommendation: - Real-time calls: tts-1 - Pre-recorded calls: tts-1-hd ``` **ElevenLabs Stability/Clarity:** ``` Stability: 0-100 (lower = more variable) Clarity: 0-100 (higher = more clear) Balanced: Stability 50, Clarity 75 Natural: Stability 30, Clarity 75 Consistent: Stability 100, Clarity 75 ``` ### Language & Accent Most providers support multiple languages: ``` OpenAI: 50+ languages ElevenLabs: 30+ languages Rime AI: 30+ languages Configuration: 1. Select language in agent settings 2. Choose appropriate voice 3. Test pronunciation 4. Save ``` ## Advanced Features ### Voice Cloning (ElevenLabs) Create custom voices: ``` Requirements: - 30-120 seconds of clear audio - No background noise - Natural speaking voice - 1-2 sentences recommended Quality: Professional, realistic Cost: Included in subscription Use Cases: Brand consistency, CEO voice ``` ### Streaming Audio Get audio as response generates: ``` Advantage: Start speaking immediately Latency: Very low (< 500ms to first word) Quality: Same as buffered Cost: Same pricing Best For: Real-time conversations ``` ### Emotion & Style Customization Available through some providers: ``` OpenAI: Fixed emotions per voice ElevenLabs: Variability control (emotion) Rime AI: Emotion settings available Use: Match tone to conversation context ``` ## Troubleshooting Try ElevenLabs for more natural sound, or adjust stability/clarity settings if using ElevenLabs. Adjust speed setting (0.5x - 4.0x depending on provider). Provide phonetic guide to agent's knowledge base, or use custom vocabulary feature if available. Switch to Rime AI (lowest cost), or reduce character count by shortening responses. Use tts-1 (not tts-1-hd) for OpenAI, or switch to Rime AI/Cartesia. ## Best Practices ### 1. Start with OpenAI ``` Best balance of: - Quality (natural sounding) - Cost (reasonable) - Languages (50+) - Simplicity (integrated with GPT) ``` ### 2. Test Multiple Voices ``` Create test agents with different voices Make sample calls Compare quality Choose best for brand ``` ### 3. Match Voice to Brand ``` Professional service: Nova or Alloy Friendly service: Shimmer or custom Technical support: Echo or Onyx Sales: Nova (warm and professional) ``` ### 4. Optimize Response Length ``` Shorter responses = Lower cost + Better experience Keep responses under 200 characters Use bullet points instead of paragraphs Remove unnecessary words ``` ### 5. Monitor Quality ``` Weekly Checks: - Listen to sample calls - Check pronunciation - Verify speed - Ensure consistency ``` ## Performance Tips ### Real-Time Optimization ``` Configuration | Latency | Quality --------------|-----------|---------- OpenAI tts-1 | 50-100ms | Good Rime AI | 50-150ms | Good Cartesia | < 50ms | Excellent ElevenLabs | 100-200ms | Excellent ``` ### Cost Optimization ``` Rank | Provider | Cost/Call -----|------------|---------- 1 | Rime AI | $0.004 2 | OpenAI | $0.006 3 | ElevenLabs | $0.012 4 | Cartesia | Enterprise ``` ### Quality Ranking ``` 1. ElevenLabs (most natural) 2. Cartesia (studio quality) 3. OpenAI (professional) 4. Rime AI (good quality) ``` ## See Also Configure voice input Compare all AI providers Create complete agents *** ## Support [OpenAI Text-to-Speech](https://platform.openai.com/docs/guides/text-to-speech) [ElevenLabs Documentation](https://elevenlabs.io/docs) Email: [callintel01@gmail.com](mailto:callintel01@gmail.com)