By installing or using LocaLens ("the app") you agree to these terms. If you don't agree, please don't use the app.
What LocaLens is
LocaLens translates text in photos, typed text and live conversation using machine-learning models that run on your device. It is provided free of charge and shows no ads (see the Privacy Policy).
Translations can be wrong
- Machine translation, text recognition and speech recognition make mistakes: words can be misheard or misread, and meaning, tone or names can be translated incorrectly or left out.
- Do not rely on LocaLens where a mistake could cause harm — for example medical instructions, dosages, allergy or safety information, legal, financial or official documents, or emergencies. Use a qualified human translator or interpreter for those.
- You are responsible for how you use translations and for checking anything important.
Your use of the app
- Use the app lawfully and respect other people's privacy — for example, ask before translating a conversation with someone, where the law requires it.
- Don't use the app to harass, deceive or harm anyone.
- Don't copy, extract, reverse-engineer or redistribute the app or the LocaLens models it downloads (including LocaLens' own fine-tuned translation models), except where the law explicitly allows it. Third-party models remain available under their own licenses, listed below.
Third-party models and software
LocaLens uses the following models and software under their licenses. Some were fine-tuned or converted by LocaLens.
| Component | Used for | License |
|---|---|---|
| Gemma 4 E2B (Google; quantized files by Unsloth, further compressed by LocaLens) | Photo and text translation | Apache 2.0 |
| Hy-MT2 (Tencent) | Translation (optional) | Apache 2.0 |
| OPUS-MT tc-big (University of Helsinki, Helsinki-NLP) — fine-tuned by LocaLens | Fast conversation translation | CC BY 4.0 |
| Parakeet TDT 0.6B v3 (NVIDIA); Brazilian Portuguese fine-tune by alexandreacff, conversion by joaorura | Speech recognition | CC BY 4.0 |
| Parakeet TDT 110M (NVIDIA) | English speech recognition | CC BY 4.0 |
| GigaAM v3 (Sber) | Russian speech recognition | MIT |
| Whisper (OpenAI) | Language detection, speech recognition for more languages | Apache 2.0 / MIT |
| SenseVoice (FunAudioLLM / Alibaba) | Chinese, Japanese, Korean speech recognition (optional) | FunASR model license |
| CAM++ (3D-Speaker) | Voice matching in conversation | Apache 2.0 |
| Supertonic 3 (Supertone) | Natural voice playback | OpenRAIL |
| PP-OCRv5 (PaddlePaddle) | Reading text in photos | Apache 2.0 |
| sherpa-onnx (k2-fsa), ONNX Runtime (Microsoft), llama.cpp | Running the models | Apache 2.0 / MIT |
No warranty
The app is provided "as is" and "as available", without warranties of any kind, including accuracy, fitness for a particular purpose or uninterrupted availability. To the fullest extent permitted by law, LocaLens and its developer are not liable for any damages arising from the use of, or inability to use, the app or its translations.
Changes
These terms may be updated; the new version will be published at this URL with a new "Last updated" date. Continuing to use the app after a change means you accept the updated terms.