Hi OpenAI Support Team,
I wanted to share some feedback regarding the performance of the GPT-4 o model when it comes to generating voice responses in Arabic. While the model does an excellent job generating text responses in Arabic, its performance in voice chat is largely incomprehensible, especially when dealing with the various Arabic dialects.
Arabic dialects vary significantly depending on the country. For instance, Moroccan Arabic is completely different from Egyptian Arabic. In daily conversations, we use these dialects rather than Classical Arabic, which is primarily reserved for writing and literature.
To illustrate this point better, I refer you to the official OpenAI blog demo featuring the GPT-4-0 model. In this demo, an OpenAI employee named “Lama” attempted to speak to the voice chat in Arabic to translate her message into Chinese for her colleague. The translation provided by the model was completely inaccurate and unrelated to her original message.
I encourage you to watch this demo to better understand the issue. Without support for Arabic dialects in voice mode, this new feature will not be very useful for Arabic-speaking users.
Thank you for considering this feedback.
Best regards,