GitHub Copilot Dictation Adds Multi-Language Support in VS Code
GitHub CopilotGitHub Copilot's built-in dictation in VS Code now runs on a multilingual on-device model, letting developers dictate in their own language instead of English only, either by following their configured language and system locale or by having the model auto-detect the spoken language. The update also adds workspace-level dictation instructions so transcripts follow project terminology and formatting conventions, plus shell-aware cleanup that preserves command syntax when dictating into the terminal. It extends the built-in dictation feature GitHub Copilot shipped in VS Code 1.131 in July 2026.
Key Takeaways
- GitHub Copilot's VS Code dictation now runs on a multilingual on-device model, letting developers speak in their own language instead of being limited to English.
- The feature can automatically detect the spoken language or fall back to a developer's system or browser locale, removing manual language switching.
- New workspace-level dictation instructions let teams tune transcript cleanup to match project-specific terminology and formatting.
- Shell-aware cleanup now preserves command syntax when dictating into the terminal instead of inserting spoken punctuation literally.
- The update builds on built-in dictation, which GitHub Copilot brought to VS Code chat, editor, and terminal in July 2026 to replace the separate VS Code Speech extension.
- Audio processing stays entirely on-device, consistent with GitHub's emphasis on local, offline voice models over cloud transcription.
Sources & Mentions
1 external resource covering this update
Dictation Goes Multilingual
GitHub Copilot's built-in, offline dictation in VS Code, introduced in July 2026 to replace the separate VS Code Speech extension, now defaults to a multilingual on-device model. Developers can set a preferred language, fall back to their system or browser locale, or let the model automatically detect which language they are speaking. All processing still happens locally on the device, keeping audio off any server, in line with how the feature launched.
Project-Aware Cleanup
Alongside language support, GitHub Copilot now lets teams add user-level or workspace-level instructions that shape how dictated speech gets cleaned up before it lands in the editor or chat. That means a team's own terminology, naming conventions, and formatting preferences can be reflected automatically, rather than requiring manual correction after every dictated message.
Shell-Aware Terminal Dictation
Dictating directly into the integrated terminal has historically been the weakest link, since spoken punctuation and phrasing do not map cleanly onto shell syntax. GitHub Copilot's update applies shell-aware cleanup specifically for terminal input, so a spoken command comes out as valid shell syntax instead of literal punctuation and filler words.
Why It Matters
Together, the three changes turn dictation from an English-first convenience into a more broadly usable input method, and address one of the more common gaps reported after the original launch: teams working in other languages, or with domain-specific vocabulary, previously got noticeably worse transcription and cleanup results.