I’d like to raise (again) an issue with the ChatGPT Android app that has been frustrating to use recently.
Previously, the app allowed users to edit transcribed voice input before sending. This was ideal because speech recognition is never perfect—especially for non-native speakers or when using technical jargon. Having the chance to quickly correct text before it was submitted saved time, improved accuracy, and made the experience consistent with the web app.
At one point, there was even a toggle in the settings so users could choose whether voice input should send instantly or allow editing first. That was the perfect compromise:
Native speakers or casual use → instant send
Non-native speakers, or those using complex/technical vocabulary → edit first, then send
For some reason, that option was removed. Now, as soon as voice input ends, the message is automatically sent—often with transcription errors that can’t be corrected in time. This has:
Degraded usability for many of us
Increased wasted iterations (and tokens)
Pushed some users (myself included) to rely on the web app instead of mobile
My questions for the OpenAI team:
1. Why was this editing option removed after it had already been implemented as a good compromise?
2. Wouldn’t allowing users to edit also provide valuable data to improve the voice recognition models themselves?
3. Can we expect this option to return—or at least an explanation for why it was removed?
This change has been a real blocker—I’ve even lost important research progress due to a single typo being sent instantly with no chance to fix it.
If others here feel the same, please add your feedback or reactions so the team sees this is not just an isolated issue.
