Concern: Retirement of Standard Voice leaves users without workable alternatives

GPT builders
, , ,
1

Hi all - please show some love if this chimes with you…

I’ve been using ChatGPT Pro heavily for months, with custom GPTs connected to Notion, Pinecone, Xero, and more. Like many others, I rely on voice-first communication because it suits the way my brain works, it’s faster than typing, and it’s safer while driving.

Here’s the problem:

Standard Voice gave me accurate GPT-5 output I could trust.

GPT Voice (Advanced) gives partial response that are still fairly shallow, often inaccurate, and although sounding more human lack the fundamental integrity of the raw model.

Custom actions don’t work at all in GPT Voice, which breaks entire workflows I’ve built around my custom GPTs.

With Standard Voice being retired on Sept 9, I’m left with only awkward workarounds (press-and-hold recording for Whisper transcription, or building my own app via Responses API).

I’m asking OpenAI to:

1. Provide a Plain Mode in GPT Voice that gives us unstyled, raw GPT-5 output. Surely possible?

2. Enable actions support in GPT Voice, so Pro users’ workflows don’t collapse.

3. Maintain accurate speech-to-text for safe and practical voice input, as standard voice function and transcription appears deliberatly degraded in the last few weeks.

Without these, users like me will be pushed into costly API workarounds or other platforms. I don’t want to migrate, but if GPT Voice remains the only option, I will use other platorms more often.

I hope OpenAI takes this seriously — otherwise the retirement of Standard Voice will be a major step backward for those of us who depend on ChatGPT every day.

4

I agree. We need a speech engine that accurately reflects the true GPT output… Not the stripped down over summarized version of the true output. Being pleasant to the ears would be a nice plus too.

5

Agreed. I have no idea why they are determined to take away one of the most useful parts of their product. Even if it’s a cost constraint - just let us pay for it.

8

Hi all, thanks for support! We got a reprieve!!!

Screenshot_20250909_162534_Discord
Screenshot_20250909_162534_Discord1080×2340 146 KB

9

I 100% agree with this. I get they want richer data from users, but our use cases require a thought partner, not shallow responses. If they want to do anything, instead they should work on improving speech to text. Like you mentioned recently seems intentionally glitchy, but there’s no amount of frustration that makes me want to deal with such a shallow model, advanced voice, in any case. This is why I left for Google. While notebookLM far less interactive, it does provide some depth.

11

Seems like a giant step backward. Isn’t technology supposed to be improving accessibility? Beyond the issue of diminished functionality is the negative impact on OpenAI users who rely on accessibility technology and tools.

12

And this is why chat threads should have the most recent update at the top… great to see this update! Hope they follow through!