Feedback on GPT-4O: Performance Issues in Non-English Languages

My name is Abderrahim, and I am reaching out to provide some feedback regarding the recent demonstrations of the GPT-4O model, specifically concerning its performance in languages other than English.

Firstly, I would like to commend your team on the remarkable quality and expressiveness of the model’s voice when responding in English. It is impressively human-like and engaging, which is a testament to the hard work and innovation that has gone into its development.

However, I have noticed a significant issue when the model operates in other languages, such as Spanish, Italian, arabic, and several others. Based on the official demos I have reviewed, the model’s pronunciation in these languages is heavily accented with an American English tone, which severely impacts the clarity and naturalness of its responses. This not only detracts from the user experience but also undermines the model’s potential when interacting in a multilingual context.

Given the model’s capabilities and potential impact, it is crucial that its performance is consistent across all supported languages. The current state, where non-English languages are rendered poorly, limits the accessibility and effectiveness of GPT-4O for a global audience.

I hope you will consider prioritizing improvements in the model’s language processing and pronunciation for non-English languages. This enhancement would undoubtedly expand the model’s usability and appeal worldwide.

Thank you for your attention to this matter. I look forward to seeing the continuous advancements and success of GPT-4O. I

Best regards,