Regression in voice dictation quality after iOS version 1.2026.230

Title

Please add a switch between real-time and context-aware voice dictation

Description

After updating the ChatGPT iOS app to version 1.2026.230, voice dictation has changed significantly.

I would like to ask OpenAI to add a setting that allows users to switch between different voice input/dictation modes, rather than forcing everyone to use the new behavior.

The previous dictation experience was much better for long, natural speech. The entire recording seemed to be processed with the context of the complete utterance. This allowed the system to revise earlier words based on what was said later, correctly determine grammatical endings, punctuation, commas, and the intended meaning of the sentence.

After the update, text appears almost immediately in the input field while I am speaking. Earlier words seem to be treated as final and are much less likely to be corrected when the rest of the sentence provides additional context.

For Russian in particular, this makes a noticeable difference. With natural, long-form speech, later words often determine the correct grammatical form or meaning of earlier words.

Feature request

Please provide a setting such as:

• Real-time dictation — transcription appears immediately while speaking.
• Context-aware dictation — record the complete utterance first, then process the entire recording using the full context before inserting the final transcription into the input field.

Ideally, users should be able to choose their preferred mode in Settings.

The previous context-aware behavior was significantly more accurate and natural for long Russian voice messages. Please consider bringing it back as an option.

Environment

• Platform: iPhone / iOS
• ChatGPT version: 1.2026.230
• Language: Russian
• The change appeared after the recent update, approximately 5 days ago.

The issue identifies the iPhone/iOS environment and requests a Settings switch between real-time and context-aware dictation, but it names no repository files, tests, or entry points. First verify whether voice dictation and iOS settings are in this repository; done would mean users can select either mode and context-aware transcription preserves full-utterance context, including for Russian.