Voice: Add language configuration for speech-to-text transcription

Status Fixed / completed
Maintainer reply None cached
Activity 4 comments · opened Mar 11, 2026 · closed Aug 17, 2026

Feature Request

The voice feature currently appears to default to English for speech-to-text transcription. There is no way to configure the transcription language, which makes it difficult to use for non-English speakers.

Problem

When speaking German (or other non-English languages) into the microphone, the transcription does not recognize the language correctly. It seems to only work reliably with English input.

Proposed Solution

Add a configurable language setting for voice transcription. Options could include:

  1. Auto-detection (preferred): Automatically detect the spoken language (Whisper supports this natively)
  2. Explicit setting: Allow users to set a preferred transcription language, e.g.:
  • Environment variable: CLAUDE_VOICE_LANGUAGE=de
  • Or in settings.json: "voiceLanguage": "de"
  • Or support a list for multilingual users: "voiceLanguages": ["en", "de"]

Use Case

I'm a German speaker using Claude Code daily. I frequently switch between English and German when communicating with Claude. The voice feature would be much more useful if it could handle both languages.

Environment

  • Platform: Linux (WSL2)
  • Claude Code version: 2.x

View original on GitHub ↗

3 Comments

Rafikooo · 5 months ago

I have a similar use case - I regularly switch between Polish and English when working with Claude Code.

My ideal solution would be separate keyboard shortcuts per language, e.g.:

  • Space - push-to-talk in primary language
  • Ctrl+Space - push-to-talk in secondary language

Currently switching requires going through /config each time, which breaks the flow. A keybinding-based approach (or at minimum a voice:cycleLanguage action bindable in keybindings.json) would make voice mode much more practical for multilingual users.

dg-hellotwin · 5 months ago

Workaround found: language setting in settings.json

The voice STT language can be configured via the language setting in ~/.claude/settings.json:

```json
{
"language": "russian"
}

  After adding this manually, /voice correctly shows Dictation language: ru and Russian speech recognition works.

  Note: I initially tried setting the language to Russian via the /config UI, and voice mode did not pick it up. I'm not sure if the issue was the capitalization or if /config didn't actually persist the value to settings.json. After editing settings.json by hand, everything worked. Worth checking whether /config reliably writes this setting.

  Tested on v2.1.76, Linux.
  
Gchapolin · 4 months ago

Portuguese (pt-BR) voice recognition is essentially non-functional

Environment: macOS, Claude Code CLI, /voice mode

Steps to reproduce:

  1. Enable voice mode with /voice
  2. Change dictation language to pt via /config
  3. Speak in Portuguese

Expected: Transcription of Portuguese speech
Actual: Either garbled English text or nothing recognizable. Examples of what was transcribed when speaking Portuguese:

  • "The cloud g u leave"
  • "Genova, Claude,"

The transcription completely fails to recognize Portuguese. Switching back to English works fine.

Would love to see proper Portuguese (pt-BR) support. Thanks!

Showing cached comments. Read the full discussion on GitHub ↗