We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·May 19·5 min read
How the voices for ChatGPT were chosen
TL;DR
Selecting high-quality voices for ChatGPT was challenging due to the large number of submissions. OpenAI collaborated with casting and directing professionals to evaluate over 400 voice samples, ultimately choosing 5 distinct voices.
✦ Why It Matters
Engineers can learn the value of expert collaboration in refining AI voice technologies for better user engagement.
Key Takeaways
- Engage industry experts for voice selection processes.
- Evaluate voice samples based on clarity and expressiveness.
- Ensure diversity in voice options for user preferences.
- Focus on user experience in AI interactions.
Full Summary
Creating engaging and natural-sounding voices for ChatGPT required a systematic approach to voice selection. OpenAI received over 400 voice submissions and engaged industry-leading casting and directing professionals to assess these samples.
The evaluation process involved narrowing down the submissions based on criteria such as clarity, expressiveness, and suitability for conversational AI. After thorough analysis, 5 distinct voices were selected to enhance user experience.
This careful selection process not only improved the quality of interactions but also ensured a diverse range of voices to cater to different user preferences. The implications of this work highlight the importance of professional input in developing AI systems that rely on human-like communication.