AI Voice: ElevenLabs vs OpenAI vs Cartesia for Realtime

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-model-families-AI-voice-cloning-models-r13a3-creators

1. What are the two main categories that modern voice generation technology has split into?

   1. Animated avatar lip-syncing versus voice dubbing for films
   2. High-fidelity offline TTS for content creation versus ultra-low-latency streaming for conversations
   3. Text-to-image conversion versus speech-to-text transcription
   4. Music generation versus narration for videos

2. Which type of voice model is best suited for creating podcast episodes and audiobooks?

   1. Real-time voice translation apps for live events
   2. OpenAI Realtime streaming models with sub-300ms latency
   3. ElevenLabs-class high-fidelity models designed for offline processing
   4. Cartesia models optimized for chatbot interactions

3. What latency target must voice models achieve to be suitable for realistic conversational agents?

   1. Under 10 seconds
   2. Under 2 seconds
   3. Under 60 seconds
   4. Under 300 milliseconds

4. Which of the following is a legal requirement before cloning someone's voice using AI?

   1. Certification in audio engineering
   2. Explicit written consent from the person whose voice is being cloned
   3. A paid subscription to the voice cloning service
   4. A government-issued ID of the voice subject

5. In many regions, voice cloning without consent is classified as what type of offense?

   1. Fraud
   2. Copyright infringement
   3. Jaywalking
   4. Petty theft

6. What limitation do current AI voice models have compared to skilled human voice actors?

   1. They cannot match the full emotional range that skilled human VO performers can deliver
   2. They cannot pronounce words starting with the letter S
   3. They cannot exceed 50 decibels in volume
   4. They cannot speak in languages other than English

7. What happens when a user interrupts a realtime AI voice agent (barge-in)?

   1. The AI immediately shuts down
   2. The AI switches to text-only mode
   3. The voice quality improves automatically
   4. The AI cannot stay perfectly on-script when users interrupt in realtime

8. Why might a content creator still need prosody coaching even when using AI voice tools?

   1. AI voices always sound robotic and require no coaching
   2. Prosody coaching is only needed for video editing, not audio
   3. AI cannot replace prosody coaching for narrative work requiring precise rhythm and tone
   4. AI has already mastered all aspects of human speech patterns

9. What capability do modern AI voice models offer regarding language and accent?

   1. Translation only, no original speech generation
   2. Only speaking in American English
   3. Mandatory accent removal for all outputs
   4. Multi-language voice synthesis with controlled accents

10. A developer building a customer service chatbot that responds vocally should choose which model family?

    1. Offline TTS models that take minutes to generate responses
    2. OpenAI Realtime or Cartesia models optimized for low-latency streaming
    3. Text-to-speech models that only output MP3 files
    4. ElevenLabs-class models designed for audiobooks

11. What is the term for AI technology that generates human speech from text input?

    1. Phoneme translation
    2. Voice synthesis
    3. Audio fingerprinting
    4. Speech recognition

12. A video producer creating a promotional video needs voice narration. Which model should they select?

    1. Any model that supports voice cloning
    2. ElevenLabs-class model optimized for high-fidelity output
    3. OpenAI Realtime streaming model for chatbot conversations
    4. Cartesia model designed for sub-300ms responses

13. What evidence should a creator maintain to demonstrate legal compliance when using voice cloning?

    1. A screenshot of the subscription payment
    2. Written consent documentation kept on file
    3. A typed agreement without signatures
    4. A social media post about the project

14. The lesson describes voice generation technology as having 'bifurcated.' What does this mean?

    1. It has combined with video generation
    2. It has been abandoned in favor of text-only AI
    3. It has split into two distinct categories with different use cases
    4. It has become twice as expensive

15. What is a key limitation when using AI voices for interactive storytelling or games?

    1. AI voices can only repeat pre-recorded phrases
    2. AI voices cannot speak any dialogue
    3. AI voices require actors to complete every sentence
    4. AI struggles to handle unpredictable user inputs and stay on narrative script