Standard Voice has been removed - badly breaks modality continuity in long-context workflows

ChatGPT
, , ,
1

Standard Voice yesterday disappeared from the normal ChatGPT browse interface for me, as well as from the Mac desktop app.

The only voice option I now get is Live Voice. I contacted OpenAI Support and, interestingly, they would not say whether this is a permanent product change. Their exact wording was:

“At this time, we cannot confirm whether the change you are seeing is permanent or provide a setting that will restore Standard Voice when it is not available on your desktop or web experience….If Standard Voice becomes available again for your desktop or web experience, it should appear among the available Voice options in ChatGPT.”

So at least from Support, there is currently no clear statement that Standard Voice has or has not been permanently retired.

This matters because Standard and Live are not interchangeable for the way some people use ChatGPT. I’m a heavy user working in long, context-rich conversations involving research, documents, analysis and development.

With Standard Voice I can move seamlessly between typing and speaking while retaining essentially the same depth, length and continuity of response.

With Live Voice, switching from text to voice halfway through a complex task is more like switching mid-task or project to a different assistant. Responses are dramatically shorter and much less substantive. Having the responses in speech be the same as I would get in text is vital if one wants to move away from the destkop, from typing, and retain full continuity.

Standard Voice is apparently still functioning on mobile apps. The capability itself has not disappeared, at least not yet, but normal desktop/browser access to it apparently has.

My request to OpenAI is simple: please keep Standard Voice available as an option alongside Live Voice. Live is useful technology, but it serves a different use case.

For users who want voice as another interface to the full ChatGPT experience, Standard Voice remains extremely valuable - I would say essential - for a lot of use cases (such as mine) where otherwise typing is the only viable option, because Live cannot replicate the interchangeable text-to-speech SVM has.

2

I might also note that Standard Voice’s component parts are still present. Dictation sends my speech into the normal ChatGPT text conversation, ChatGPT generates the full response, and Read Aloud speaks that response in the same Standard-style voice.

Standard Voice simply automates those steps into a continuous hands-free conversation (and also, usefully, allows for me to interrupt a long answer by speaking).

If OpenAI is continuing to support all those components, it is difficult to understand why the integrated Standard Voice workflow should be removed and users forced to press two buttons on every turn to partially recreate the interface manually. That’s a clear regression.