- On-device translation using Mozilla’s translation models
- Transliteration of non-latin script
- OCR (Optical Character Recognition) for translating text in images
- Automatic language detection
- Image translation overlay that preserves original formatting
- Support for multiple language pairs
- No internet required for translation once models are downloaded
- All translation happens locally
I’m wondering how ff models compare to smth more modern, like the recent translateGemma stuff (given 12b models reportedly run fine on phones nowadays)