BUG - Gemini requests are getting timed out very quickly.

[avinashkanaujiya]

Is there a timeout enforced by Voiceink itself? Gemini requests—especially 2.5 Flash model requests—are getting timed out very quickly. Maybe Voiceink is applying this timeout, or it could be an issue on the Gemini side.

I’m fine with transcriptions taking 20-30 seconds to complete because the prompts I work with are very complex, so they naturally require more time. Perhaps consider increasing the request timeout for thinking models. If it’s not Voiceink causing this, I’m unsure why my requests keep timing out so quickly. This issue has been occurring since July 27th or 28th.

[avinashkanaujiya]

The error doesn't seem to be coming from the timeout itself because it occurs very quickly. It's not even taking 10 seconds for the error to hit.

Something else might be causing this issue.

[michael-sumner]

I can confirm that I'm receiving this issue for gemini-2.5-pro also. Removing the "AI provider integration" API key solves the issue temporarily.

[avinashkanaujiya]

@michael-sumner You mean like removing the AI enhancement altogether?

[michael-sumner]

@avinashkanaujiya yes try seeing if you can remove the API key and then if it works?

[avinashkanaujiya]

I already tested it and I know why it's happening: voicing had a default limit of 10 seconds, and now it's about 30. If any prompt is complicated, or Gemini models tend to get stuck and take more than thirty seconds to complete, it will hit the timeout.

It will show up as a custom error: "The request timed out."

[michael-sumner]

@avinashkanaujiya looks like if you also set the "AI Provider" to "Custom" (Even if it's empty) then fixes the "long transcribing..." issue.