I’m raising this as an accessibility issue, not simply a UI preference.
I use ChatGPT primarily through a very specific workflow: I speak my messages using ChatGPT’s voice-to-text (VTT), receive the normal full written response, and then use Read Aloud to listen to that response.
This is important for me because I’m neurodivergent and have difficulty both typing long messages and processing long written responses. Voice Mode is not an equivalent replacement. I want to formulate my own long messages at my own pace, receive the complete written answer, retain that text in the conversation, and then listen to it.
Other users have already raised essentially the same accessibility issue after the dedicated Read Aloud button was removed from the response toolbar to the overflow bar, RIGHT NEXT TO ‘retry’.
OpenAI Support has acknowledged that feedback, including the fact that Voice Mode does not replace this workflow, but the button still has not been restored.
The particularly frustrating thing is that the response toolbar now has plenty of empty space where the Read Aloud button used to be.
Now another update has made the input side of the same workflow significantly less accessible.
As of today, VTT on iOS displays the transcription live in the message composer while I am still speaking. I’ll attach a screenshot showing this.
For some users this may be useful. For me, and potentially other people with ADHD, autism, dyslexia or other conditions affecting attention/language processing, it creates a serious problem.
The reason I used ChatGPT VTT instead of ordinary dictation was precisely because I did not have to watch and monitor text appearing while I spoke.
When transcription errors were not displayed to me during speech, they generally did not matter. I could simply speak naturally, finish my thought and send it. Minor transcription errors could be understood from context.
Now, moving text is appearing directly in front of me while I speak. My attention is automatically drawn to it. I see transcription errors happening in real time and start monitoring or correcting them. Within minutes of receiving this update, I noticed that I was already speaking differently because I was watching the transcript rather than concentrating solely on what I wanted to say.
Telling users to simply “not look” is not a meaningful accessibility solution when dynamically changing text is being displayed in the interface they are actively using.
This effectively introduces into VTT the exact feature that made ordinary dictation inaccessible to me in the first place.
I’m not asking for live transcription to be removed for users who find it helpful. Please simply provide an option such as:
Show transcription while recording: On / Off
With it disabled, VTT could behave as it did previously, allowing the user to speak without seeing the transcript until recording has finished.
More broadly, please consider this workflow as a whole when making UI changes:
VTT → complete written response → one-tap Read Aloud
For some of us, these are not three unrelated convenience features. Together they form an accessibility interface that allows us to use ChatGPT effectively.
It is therefore particularly frustrating to see another update alter that workflow after multiple rounds of accessibility feedback about the removal of the Read Aloud button have already been acknowledged, while the Read Aloud button itself still has not returned to an otherwise largely empty toolbar.
Please restore the one-tap Read Aloud button to the toolbar and provide an option to disable live VTT transcription while recording.
