Skip to content

Promote to production: voice-to-text language selector + larger model - #81

Merged
slaveofcode merged 2 commits into
mainfrom
develop
Aug 1, 2026
Merged

Promote to production: voice-to-text language selector + larger model#81
slaveofcode merged 2 commits into
mainfrom
develop

Conversation

@slaveofcode

Copy link
Copy Markdown
Owner

Adds a language selector (fixes Bahasa auto-detect misfires) and an opt-in whisper-small multilingual model.

🤖 Generated with Claude Code

Kresna and others added 2 commits August 1, 2026 17:44
…r Bahasa

Bahasa (and other non-English) accuracy was poor because Whisper was auto-
detecting the language (often mis-guessing on short clips) on the small
whisper-base model.

- Add a Language selector (default Auto-detect) shown for multilingual models;
  pass the chosen language + task:'transcribe' to Whisper. Naming the language
  is the biggest accuracy win at zero download cost. Indonesian/Malay/Javanese/
  Sundanese listed first, plus major world languages.
- Add 'Multilingual · Better' (onnx-community/whisper-small) as an opt-in model —
  markedly better for non-English, larger one-time download. whisper-base stays
  the default multilingual.
- Engine: SttModelId gains whisper-small; transcribe() takes an optional
  { language } (applied only for multilingual models).

545 tests · lint clean · build green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
feat(voice-to-text): language selector + larger model for Bahasa
@slaveofcode
slaveofcode merged commit d3b5c15 into main Aug 1, 2026
2 of 3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant