I also suffer a lot from this downgrade!
2 Likes
Any solution to reverse this?
2 Likes
I fully agree with the above, it’s totally destroyed my primary usecase. The lack of being able to re-send or save the audio for transcription later is very very frustrating, it seems like at least 1 in 5 audio clips are just lost and it’s hard to remember what one has said. This is compounded with clips that are longer so around 4 minutes or more.
2 Likes
Does anyone know if Gemini works as well as GPT used to? Or Claude?
In this sense I mean, on the subject of voice dictation
1 Like
Ich möchte mich ebenfalls anschließen. Seit der neuen Änderung der Sprachfunktion fühle ich mich massiv eingeschränkt. Früher konnte man die Aufnahme überprüfen, gegebenenfalls neu starten oder verbessern. Jetzt wird alles sofort abgeschickt, selbst wenn die Verbindung kurz wackelt oder man sich verspricht. Ich habe dadurch schon unzählige Nachrichten verloren – zum Teil sehr lange, sehr persönliche Texte – und musste sie neu einsprechen. Das ist extrem frustrierend und kostet mich Kraft, die ich nicht habe.
Ich habe mehrfach versucht, damit klarzukommen – aber für mich ist das nicht mehr tragbar. Aus diesem Grund werde ich mein Abo kündigen. So, wie es jetzt ist, ist die Sprachfunktion für mich nicht mehr nutzbar. Ich bitte das OpenAI-Team dringend, die alte Funktion wiederherzustellen oder zumindest eine Option anzubieten, wie man die Aufnahme bewusst absenden kann, statt dass sie automatisch losgeschickt wird.
1 Like
Yup, it has been fixed quite fast. Glad they took care of it. Plus, now ChatGPT on Windows also features the “dictate” function. (Maybe it’s been there before, and I just haven’t noticed yet lol)
Is this only on desktop? The mobile [Android] app I’m currently running, v1.2025.091 is still plagued with this issue.
Hopefully the change reverts on all platforms.
Interesting. No, for me, it was fixed on the mobile version as soon as it was fixed for @javifeeble (Apr, 3rd). I thought it was already reverted for everyone already. For me, on my Android, I am running version 1.2025.091 (16)
Also, I am based in the EU.
Workaround : uninstall chatgpt, find old APK (I used uptodown to find the version, before April 1st, I think the one on March 27th is the one I downloaded) and installed the old version.
The old voice function worked so much better.
1 Like
Hi! So for me (using GrapheneOS) the voice dictation STILL automatically sends.
I still think that this is a major downgrade.
1 Like
This is a disaster. It barely understands any technical/business terms and almost always get them wrong and I always have to tweak it. Also if you talk for more then 2 minutes it often loses your thought train completely without recovery if any network errors occurs or if it hallucinates and returns some transcribed nonsense.
Previously I was going a few sentences at a time, then fixing all errors it made, and then dictate next sentence again.
Please turn auto send off!!!
I’m on Google play beta version
I appreciate the new one-tap voice-note send — it’s really slick when I only need to say a quick question or prompt. That said, I frequently record longer dictations in multiple chunks so I can review, refine and combine them before sending. Having the transcript auto-submit straight away now breaks that workflow.
Could we please have an “Edit before send” option as a secondary action (e.g. long-press or a small toggle in Settings)? Leaving the current tap-to-send as the default is totally fine, but I’d love a quick way to route dictations into the text box for review/editing before they go through. Otherwise I’m forced to send, then manually fix typos in the chat—feels like a hack more than a solution.
Thanks for considering this enhancement! 
1 Like
The issue is still present.
I’m using the ChatGPT Plus Android app (latest version), and for several days now, the voice transcription feature has been failing. When I record voice input, the recording does not transcribe and is lost entirely.
And I also very much prefer the option to edit my recording before it gets send. That also allows me to make one message with multiple recordings. Please revert it or fix it, thank you! A lot of time/energy is being lost into failed transcriptions.
1 Like
I wanted to follow up on this post. Today, as of May 6th (at least for me, on Galaxy S22), the speech to text function returned to it’s original version for me.
This has been an immense relief and hope to God that this isn’t some A/B testing fluke and it’s here to actually stay this time.
1 Like
Totally agree with the previous message
Yeah totally agree.
It was super helpful to speak out loud.
Hey the function is back.
But still has an major flaw.
Sometimes it converts the speech and end in a black chat with no chance to send the message.
Nothing converted to text
It’s on android. Red mi 10
Pls have a check for it
2 Likes
Same here: The dictation feature is back, but it doesn’t work in most cases. Often, nothing gets converted to text, and I also end up in a black chat screen with no way to send the message. I’m using an Android device with the following ChatGPT app version:
ChatGPT version: 1.2025.119
Android 14
1 Like
Same issue on Android App.
Dictation in the chat (not the voice chat) fails 8 out 10 times, both on working wifi at home and using mobile network out and about.
You dictate, it says “converting to text” then you get a blank text box except for “Ask me anything” with no text and no option to do anything. Restarting the app does nothing.
As a paying user, I’m quite annoyed by this.
Dear OpenAI Team,
I’m using ChatGPT on multiple devices – iPhone, iPad, Mac, and in the browser. The voice-to-text functionality still works beautifully on the iPhone, on the Mac app, and also in the browser.
However, on the iPad version, the situation is exactly as described: after dictating, the text is immediately sent as a prompt without giving the chance to review or edit it first.
I sincerely hope this change will be reversed on iPad as well.
It was such a great feature to be able to edit the text after dictation – to correct typos on names, or just include something I had forgotten initially. That made for a wonderful and efficient workflow:
Sitting in the garden, dictating into the iPad, reviewing the text, adding a snippet, dictating again, editing — and only then sending it as a prompt.
Please, please, please OpenAI – bring that functionality back to the iPad.
It truly made the experience so much better.
Thank you very much!
Best regards,
Michael Grauer