Hi there! 👋
I understand you're experiencing issues with the OpenAI Realtime model's voice speed and pitch when used over the phone. While I don't have a guaranteed fix for this specific problem, I can definitely share what I know about this issue and some potential troubleshooting steps.
Based on the information I've found, this is a known issue with the OpenAI Realtime model. Here's what might be happening:
First, it's important to note that the OpenAI Realtime model is still in beta mode, so some hiccups are expected. According to [user discussions](
https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1336440478727667844), you're not alone in experiencing these issues!
One key insight: the speed setting may not actually work with the Realtime model. As mentioned in [this thread](
https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339276020389253181), "you cannot control the speed of Realtime model and these fields are left over there for backward comparability." This explains why adjusting the speed to 4 doesn't help.
Some potential approaches to try:
1. **Consider using a different model for production**: Since the Realtime model is in beta, you might want to use GPT-4o for production and keep Realtime for development/testing until these issues are resolved.
2. **Check your call logs**: Look for any model switching during calls, which could affect voice characteristics.
3. **Review your configuration settings**: Make sure your assistant's language settings are properly configured, as missing language settings for the voice model have been identified as a potential issue.
4. **Try different voice providers**: Some users have had success switching to different voice providers when experiencing similar issues.
Is there anything specific about these suggestions you'd like me to elaborate on, or do you have other questions about using the OpenAI Realtime model?
Sources: