I'm encountering a frustrating bug with the Google Gemini app: when I use its in-app microphone for dictation, it simultaneously starts speaking aloud from a previous, unrelated conversation while still recording my current input. This issue is specific to the Gemini app's dictation feature, as my Samsung keyboard's dictation works correctly. It appears to be a problem where Gemini mistakenly triggers its text-to-speech for past content during live voice input, despite me not asking it to read anything aloud.
I've already tried extensive troubleshooting, including clearing the app's cache and data, verifying permissions, checking voice settings within Gemini, updating the app, and restarting my phone, all without success. This suggests a deeper app-specific bug. For the report, my phone model is Samsung Galaxy S25, and I can provide my Android and Gemini app versions if needed.
Start a new chat: Copy relevant information from the previous conversation and restart the chat to give Gemini a clean slate.
Use explicit instructions: Include phrases like "Continue based on our last messages" or summarize the current state of your project to reinforce the context.
Use structured prompts: Ask Gemini to summarize the previous conversation and use that summary as a reference for new queries.
Reduce input size: If you are providing very long prompts or lists, try shortening them to stay within the model's token limit.
Hard refresh and Incognito/Private browsing: These steps can help rule out browser extensions or cache issues.
Clicking "redo" or editing the prompt: Clicking the redo button or editing the prompt after sending it might resolve the issue.
Important Note: While Gemini has introduced a feature to recall past conversations for improved responses, this seems to be a separate issue related to unexpected repetition within a single session, not the intended memory feature.
