ChatGPT

**Voice recognition is practically unusable** ChatGPT’s voice recognition has become so unreliable that it actively prevents me from working. It does not merely misunderstand individual words. It invents words I never said, expands short terms into completely different expressions, switches languages without reason, and sometimes changes the entire meaning of a sentence. For example, I clearly dictated the word “sub,” but the system changed it to “subdomains.” On another occasion, I said, “Beautiful, great, it works,” and the transcription produced unrelated nonsense. These are not minor spelling mistakes. The system is generating content that was never spoken. This is especially dangerous when dictating technical commands, directory names, server paths, configuration instructions, or deletion commands. One invented word can completely change the requested operation and potentially cause serious damage. I already have Slovak selected as the preferred language. The microphone is working correctly, and I speak clearly. Repeating every second or third sentence several times is not an acceptable solution for a paid service. Voice recognition should transcribe what the user actually says. It should not guess, rewrite, “improve,” translate, or hallucinate additional words. In its current state, this feature is unreliable, frustrating, and unsuitable for any serious work.

Not working after login

Voice recognition through the microphone is not working, at approximately the same time as yesterday.

The chat gpt page, when switching to microphone mode, i.e. when it is supposed to listen to what I dictate, crashes with the following error: Content failed to load

Stránka chat gpt pri prepnutí do režimu mikrofón čiže keď má počúvať to čo diktujem stránka padne do nasledujúce poruchy: Content failed to load -------------- The chat gpt page, when switching to microphone mode, i.e. when it is supposed to listen to what I dictate, crashes with the following error: Content failed to load