Ah support is right - there was an update to the chat gpt skill a few weeks ago for doing images and such for gpt-4o. check it out - there's a conversation at the bottom with nink using it

GPT-4o nows inludes audio input. Can this be utlised in the ChatGPT Skill? For my robot, the latency for verbal communcation is much higher than for Image description. I assume there are two reasons for this. Firstly, the ChatGPT Skill is converting my speech to text, and secondly, when the BingSpeech Recognition Skill starts recording, it waits until the maximum recording length expires before sending the text to ChatGPT. So if the user says 'hello', ChatGPT takes several seconds to respond. I think this skill used to detect when the person stopped talking which would be more efficient. I tested GPT-4o in real time via my account, the response was immediate, and as soon as I stopped talking. This is what I'd love to achieve with my robots.

Sounds cool. Id still like having my robots at least sound mechanical when they speak ( like  a Cylon)