Problem
The Whisper model used for voice transcription is hardcoded. Switching to a multilingual model (e.g. large-v3 or a language-specific variant) requires manually replacing the model file — a fragile workaround with no UI support.
Proposed solution
Add a telegram.whisperModel key to settings.json:
{
"telegram": {
"whisperModel": "base"
}
}Sensible values: tiny, base, small, medium, large-v3, large-v3-turbo.
Surface the option in the Telegram setup wizard with a brief explanation of size vs. accuracy tradeoffs. Default remains the current hardcoded model for backwards compatibility.
Source
Requested by @miraculix95 and @zodo in #14 (Mega-Post).
Problem
The Whisper model used for voice transcription is hardcoded. Switching to a multilingual model (e.g.
large-v3or a language-specific variant) requires manually replacing the model file — a fragile workaround with no UI support.Proposed solution
Add a
telegram.whisperModelkey tosettings.json:{ "telegram": { "whisperModel": "base" } }Sensible values:
tiny,base,small,medium,large-v3,large-v3-turbo.Surface the option in the Telegram setup wizard with a brief explanation of size vs. accuracy tradeoffs. Default remains the current hardcoded model for backwards compatibility.
Source
Requested by @miraculix95 and @zodo in #14 (Mega-Post).