In the latest iOS version of the ChatGPT app, the voice recording feature now requires dragging and holding to record. This replaces the previous lock to record option which worked well for hands free input. I’m not sure if this is a bug or a design change, but it makes the feature much harder to use.
I use voice input often to capture longer thoughts or meeting notes during work. Holding the screen the entire time is physically uncomfortable, especially for users like me who already try to limit thumb use due to strain. The old system also allowed for reviewing the transcript before sending, which is now much less intuitive.
If possible, please consider reintroducing the lock to record option or making it available as a setting. It made a big difference in daily usability.
Absolutely agree. There has also been weird sporadic updates on the app regarding the dictation feature - one day it allows you to hold and lock, another day it’s back to what it was and now this.
This feature alone is almost enough to make me switch to another LLM like Gemini. The dictation as ccuracy on CGPT is uncanny and probably the main reason why I’m sticking with it.
It seems like we are in a test group. This “feature” is horrible. I am now only using the apple dictation in-build option, but it’s nothing like whisper. I keep leaving negative feedback to make sure the devs will get my message…
So frustrating
One of the most valuable parts of the previous voice input system was that it allowed hands-free recording while still being able to scroll through the screen. This meant I could reference earlier messages, health data, or specific instructions as I spoke. I often rely on that functionality when reviewing medical results or working through detailed prompts, and being able to scroll while dictating made my input more accurate and thoughtful.
I am so frustrated as this “feature” removes that entirely.
Now I can only see what’s on the screen when I start recording, and I’m unable to scroll while holding the button.
This completely breaks the dynamic of how I use the app, not just for casual input but for serious, focused work.
It also causes real physical discomfort!
My thumb hurts after using the feature, and to be honest, I’ve now stopped using it entirely. I’ve switched to Apple’s built-in dictation as a workaround, but that system doesn’t match Whisper’s quality, and it also prevents scrolling while dictating, so the same problem remains.
I agree it has messed up my entire flow. We were able to transcribe, see what was transcribed then add more if needed. This drag feature is trash and doesn’t let you review it before it sends it off for analysis . Bring the previous version back and leave it alone please and thank you.
I completely agree. It’s awful having to keep holding the screen. Imagine you’re wearing headphones and want to send a long voice message—it’s just not practical to keep pressing the screen the whole time.
If at least there was a “lock” feature so you could record without holding—like WhatsApp does—it would be better. But honestly, the old version was 100% better. Just tapping once and letting the audio run freely was so much more convenient.
Please bring the old behavior back. This update just doesn’t work well. It’s actually making me consider switching to other AI chat apps just because of this.
Also, the real-time voice mode doesn’t solve the issue—I sometimes just want to send a long audio message, not have a real-time conversation with the AI.
And another problem: there are frequent accidents where I accidentally lift my finger and it stops recording, breaking the whole train of thought and wasting time and tokens unnecessarily. It leads to more time spent, more AI usage, more energy, all because of a small accidental gesture. It’s frustrating and counterproductive.
PS
If anyone knows how to revert to the previous version on iPhone or a setting to disable this new behavior, please let me know.
Couldn’t agree more. I don’t understand what the rationale was here and why they don’t allow users to activate/deactivate this by choice via settings rather than forcing us to use it this way. I’m paying the same monthly premium for much less convenience, fix this or you’ll lose people. I’m starting to look for a better app but first need to transfer all ChatGPT’s knowledge about me to another app so that we can continue from where we left off. Anyone know the best way to do that?
I think, guys, what we should do is flood OpenAI with feedback - I keep giving the thumbs down in the browser, choosing ‘Other’ as the feedback type, and I just express my, let’s say… gently, disappointment.
I don’t want to have to switch to another model, but if they don’t revert this they’ve killed the only reason why I’ve been 100% loyal to ChatGPT
Effortless access to whisper, this is a disability issue. Is this even legal in Europe? I bet they don’t have it there.
This is not an app level issue. It only affects certain accounts. Logging on to a different account on same app same device doesnt give the issue.
Lers hope good sense dawns and they rollback this else I’m cancelling subscription and quitting.
I absolutely hate the new update. I also used this feature to take notes during meetings and etc. The drag and hold is very inconvenient and no time to edit prior to sending. Old layout please.
I’ve just discovered one more annoying thing about that “update”:
Whisper could automatically detect the spoken language, even when switching between multiple languages in a single recording. I used this to practice Arabic, check pronunciation, and mix in English or Polish.
Now, transcription is locked to the app’s default language, regardless of what I actually say. This breaks the workflow completely and makes language learning through voice input nearly impossible. I used to rely on this feature daily, and its removal is a major loss for Imultilingual users.
I was a heavy user of the previous mic feature, as the live voice obnoxiously interrupts you everytime you want to use it.
Now, is worse, I have to actually hold my thumb down while I speak, and I am left handed which makes it super awkward.
Anyone know what other models have a more friendly and easier to use voice feature?
So, from my experience, Whisper (by OpenAI) is the best voice dictation option. It catches everything you say and produces grammatically correct text, it used to switch between languages in one input, and catched even “ummm”s…
Grok and Gemini transcribe the text live on the screen when you speak (which slows speaking down in my opinion as you follow along the text…). But! With these, you don’t have to manually press your finger to the screen, which is good (ekhem @OpenAI_Support ). Their dictation behaves quite similarly to Apple’s built-in dictation feature. Solid but lack of this “whisper magic”. You can also review the text on both Gemini and Grok before sending. As for Claude, you can use the mic button to record and transcribe your dictation, but it won’t let you review it beforehand.
In my opinion, all three of them are solid dictation tools, but they do make mistakes. As I am not a native English speaker, I find they make a lot more mistakes for me – I don’t have exact numbers, but significantly more than Whisper. Whisper somehow grasps the context and can understand what I’m saying, even if my pronunciation isn’t perfect.
I really miss whisper. I’ve quit using it on a daily basis….
