I have a custom GPT where function calling is working in all scenarios except when using advanced voice mode. This is a regression where previously with the legacy voice mode the custom GPT was capable of executing functions and generating a response without issue.
Anyone else run into this? Any suggested workarounds?
This GPT is integrated at a critical point in our business and our end users are typically speaking to the custom GPT so this problem is hugely impactful.
Same here… That’s really a big limitation.
Any updates on this, will we get this feature again? In https://help.openai.com/en/articles/8400625-voice-mode-faq it does not sound like that (“Custom actions in GPTs are not available in Voice mode“).
This is a MAJOR issue that I am shocked is not being addressed by more people. It appears another thread on the topic was closed a day after it was opened. @OpenAI_Support This NEEDS to be addressed. Actions do not work in advanced voice mode with custom GPTs and even worse the model hallucinates and provide bogus information when you continue to prompt it to use an action. This is a disaster!
Why is this issue not sorted already. I have been using ChatGPT since the day it was released and using custom GPT to perform actions is a big part of my daily workflow. I can accept the Advanced Voice not working if i at least still had the black dot. Please fix this!!!
@OpenAI_Support What is going on here?!?!?! Why are we not getting any attention to this matter. It has completely slowed down our operations as we now have to chat with the GPT instead of being able to interact through voice mode. This has to be impacting 1000’s of users. PLEASE respond.
I am still blocked and do not have any workarounds. A workaround is the worst outcome but at least renders the solution usable.
Right now the Custom GPT has been unusable for almost 2 months. This is making me reconsider if OpenAi is truly enterprise grade. No enterprise solution would ignore a major regression for 2 months without at least acknowledging and providing an estimated resolution date. This feature is critical enough that we are exploring other platforms as Custom GPTs do not feel Enterprise Grade.
Don’t build anything critical using OpenAi’s api calls. Now if the things don’t go through your business is f**ed. Have learned this the hard way!
My company did a massive project spending money on an outside consultant to develop a business process that relied on advanced voice at a critical step. After much attention, it worked and we rolled it out to all the people in the company. Many were AI skeptics, but eventually adopted the new practice then Open AI pulled the rug out from under us and remove the advanced voice functionality. I have learned from this is you can’t rely on the Open AI tools for business processes and incorporate that into Real organization change. Perhaps this is because open AI is not a real company and this nonprofit aspect makes them think differently about customer service. I wish I had built that new tool on a different platform and that is the direction I’m sure we’ll go in the future. Beware of Open AI. It is not a reliable tool for real use in business process processes. It might be a fun consumer product, but can’t be relied on when you’re bringing AI into business. Try one of the other platforms, especially if you have AI skeptic employees who will feel vindicated when a new tool works for a month or two and then permanently fails.
We’re using a Custom GPT not the APIs and learning the hard way 
@OpenAI_Support PLEASE pay attention to this!
Everyone reading this needs to start commenting to get their attention.
@SamAltman you probably want to see this
@OpenAI_Support any insight into expected resolution date?? Should I be moving off the Open AI Platform to a tool that better supports function calling with voice mode or is this something you have prioritized to resolve soon?
Function calling in Advanced Voice Mode is still broken.
No API calls, no webhooks, no actions — nothing works while talking.
Nobody cares about this feature? Or am I one of the few crazy people actually trying to use voice mode for real workflows?
Cheers
Also a bit disappointed this does not work, seems like voice chat is a big selling point for using ChatGPT over providers like Google Gemini 3. I want to be able to interact with things through voice rather than text.
Bumping this. “Advanced Voice” does not work with a Custom GPT I’ve made, with actions under the Open API 3.X framework. We would love a solution here.
That said - the non-”Advanced Voice” voice does work as expected. Though it’s less-human like.
I encountered a similar issue. Advanced Voice Mode seems to manage the response pipeline differently, which can cause function calls to fail even when they work perfectly in text mode. What helped me was implementing a stricter schema and moving the function-call trigger into the system instructions. This situation is definitely a regression, so I recommend reporting it to OpenAI as well. In the meantime, the only effective workaround is to keep users on the legacy voice mode or route voice input through text first.