[Feature Request] Custom vocabulary support for voice dictation to improve technical acronym and accent handling

Bug Description
Voice dictation: no custom vocabulary, and accent/acronym handling struggles with technical jargon

Voice dictation frequently mis-transcribes domain acronyms and short technical words, and seems to struggle with a New Zealand accent (vowel sounds in particular). Because transcription happens before Claude sees the text, project context (CLAUDE.md, memory, branch names) can't correct it — the acronym is already gone by the time Claude reads it.

Real examples from a single session:

• "send to a PR" → "send to appear" ("PR" lost as a word entirely)
• "Cool, enter a new worktree, please" → "Call Inter Anu Work Tree, please" (vowel-heavy phrase, NZ accent)

Requests:

1. A custom vocabulary / dictionary file (e.g. .claude/voice-dictionary.json) where a project can declare its acronyms and terms — "PR", "OTel", "ECS", "OpenIddict", "worktree", "RDS", etc.
2. Spoken→written replacements (e.g. "PR" → "pull request", or just reliably keeping "PR" as a token).
3. Better accent robustness, or a way to bias the model — NZ/Australian vowels get mangled more than the pre-tuned coding vocabulary suggests.

Even just (1) would solve most of the pain for jargon-heavy projects.

Environment Info

• Platform: linux
• Terminal: ghostty
• Version: 2.1.195
• Feedback ID: d9621f6d-da12-4768-ac1c-0f49cee488f8

Errors

[]